TAO Sparse4D Model Card

Model Overview

Description:

TAO Sparse4D is an advanced three-dimensional (3D) multi-camera detection and tracking network. It generates precise 3D bounding boxes and tracking IDs for a diverse set of objects across multiple camera views. The model was developed by NVIDIA as a part of the TAO model family. The provided model is pre-trained on the Multi-Target Multi-Camera (MTMC) Tracking 2025 subset from the NVIDIA PhysicalAI-SmartSpaces dataset also utilized for the 2025 AI City Challenge. This model is ready for commercial use.

The model in this card was trained & evaluated on the following warehouse-centric classes: Person, Fourier_GR1_T2_Humanoid, Agility_Digit_Humanoid, Nova_Carter, Transporter and Forklift.

License/Terms of Use:

GOVERNING TERMS: Use of this model is governed by the NVIDIA Open Model License.

Deployment Geography:

Global\n

Use Case:

TAO Sparse4D is designed for 3D multi-camera object detection and tracking in indoor environments like warehouses and logistics facilities. The model detects & tracks objects across multiple camera views for applications including warehouse automation, safety and workflow optimization in industrial settings, providing spatial understanding of a scene.

Release Date:

v1.0 Model - NGC - 06/13/2025
\nv1.1 Model - NGC - 11/30/2025
\nv2.0 Model - NGC - 01/23/2026 via NGC
\nv2.1 Model - NGC - 02/27/2026 via NGC
\nv2.2 Model (ResNet50) - NGC - 05/12/2026 via NGC
\nv2.2 Model (ResNet101) - NGC - 05/12/2026 via NGC

References

Model Architecture

Architecture Type: Convolution Neural Network (CNN) based backbone with Transformer based decoder layers.

Network Architecture: ResNet101 backbone with Transformer based decoder layers.

Number of model parameters: 7.8 × 10⁷ parameters

This work leverages the Sparse4D v3 model & is tailored for indoor environments like warehouses with static camera setups. Sparse4D is a query based technique which involves sampling sparse features for better computational efficiency compared to other dense 3D detection & tracking based techniques. The architecture features a ResNet101 backbone and processes time-synchronized frames from multiple cameras. Key components include the backbone, a Feature Pyramid Network (FPN), and multiple decoder layers incorporating Multi-Scale Deformable Aggregation blocks (featuring key-point generation, feature sampling, visibility net module & embedding generation) along with refinement & classification layers. The model is trained on regression, classification & ID losses. The model also consists of an Instance Bank module utilized for tracking ID assignment & management. Along with the 3D bounding box information & object ID, Sparse4D also outputs an instance feature containing high-dimensional semantics from the image encoder. The training algorithm optimizes the network to minimize the classification loss, regression loss & ID loss.

Input:

Sparse4D is trained on RGB images, camera calibration files, ground truth 3D bounding boxes & object ids, and optional depth maps. Since performing 3D multi-camera detection and tracking across large regions like warehouses with high-density camera setups is computationally expensive, we partition large regions into overlapping groups called Bird's Eye View (BEV) group. Each BEV group contains multiple cameras and serves as the fundamental training unit for Sparse4D.

Input Type(s): Each BEV group will have cameras with RGB images, Camera calibration file, Ground Truth with 3D bounding boxes & Optional Depth Maps available via HuggingFace or generated synthetically by Nvidia Isaac Sim Replicator.

Input Format: Red, Green, Blue (RGB) images stored in raw png/jpg/hdf5 and Depth Maps stored in png/hdf5. Can support input resolution of 3 x 1080 x 1920 for both RGB & Depth Maps. Data preprocessing is required for grouping images & depth maps from multiple cameras to appropriate BEV groups. This can be done via the TAO Data Service.

Input Parameters: Multiple dimensions. See below for detailed model input shapes.

Other Properties Related to Input: 3 x 1080 x 1920 (C x H x W) resolution images for both RGB & Depth Maps. Data Pre-Processing needed via TAO Data Services. No alpha channel required.

Note that depth maps are optional and not required for evaluation or inference. The raw model inputs are as follows:

\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n

tr:nth-child(odd)]:bg-surface-sunken\">

Dimension	Description
`B`	Batch size
`C`	Number of channels
`N`	Number of cameras
`H`	Image height
`W`	Image width
`Q`	Number of queries
`M`	Number of output boxes
`E`	Number of instance features

\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n

tr:nth-child(odd)]:bg-surface-sunken\">

Input Name	Type	Shape	ResNet101 Shape	Description
`img`	List[Tensor]	(B, N, C, H, W)	(1, N, 3, 540, 960)	Input image tensor
`projection_mat`	List[Tensor]	(B, N, 4, 4)	(1, 10, 4, 4)	List of projection matrices
`image_wh`	List[Tensor]	(B, N, 2)	(1, 10, 2)	List of image width and height
`input_cached_feature`	List[Tensor]	(B, 600, 256)	(1, 600, 256)	List of cached features
`input_cached_anchor`	List[Tensor]	(B, 600, 11)	(1, 600, 11)	List of cached anchor
`prev_exists`	List[Tensor]	(B)	(1)	Indicates if previous frame exists or not
`interval_mask`	List[Tensor]	(Bx1x1)	(1x1x1)	Boolean to describe the interval mask

No. of cameras is dynamic in the model.
Model is initialized with zero tensors for the input_cached_feature, input_cached_anchor, prev_exists & interval_mask at the first frame. These tensors will be updated automatically via instance bank from the second frame & onwards.

Output:

Output Type(s): Tensors consisting of bounding boxes in 3D , object confidence scores, classes, class confidence scores, tracking object ID & instance features.

Output Format: List of Tensors.

Output Parameters: Multiple dimensions. See below for detailed model output shapes.

Other Properties Related to Output: Please see the details below.

The final output is a list with length of batch size after post-processing. Each element is a dictionary with the following keys:

\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n

tr:nth-child(odd)]:bg-surface-sunken\">

Output Name	Type	Shape	ResNet Shape	Description
`boxes_3d`	List[Tensor]	(B, M, 10)	(1, 600, 10)	List of 3D boxes: `(x, y, z, width, length, height, yaw, velocity_x, velocity_y, velocity_z)` in OV coordinates
`scores_3d`	List[Tensor]	(B, M)	(1, 600)	List of object confidence scores (classification scores * centerness score)
`labels_3d`	List[Tensor]	(B, M)	(1, 600)	List of class labels
`cls_scores`	List[Tensor]	(B, M)	(1, 600)	List of classification scores
`instance_ids`	List[Tensor]	(B, M)	(1, 600)	List of instance IDs
`instance_feats`	List[Tensor]	(B, M, E)	(1, 600, 256)	List of instance features

where:

x, y, z - 3D coordinates of the bounding-box centroid in the world coordinate system which is in meters.
width, length, height - Box dimensions in meters along its x (width), y (length) and z (height) axes of the object-centered coordinate system, with the origin at the centroid.
yaw - Euler angle in radians about the y-axis of the object-centered coordinate system defining the box’s heading in the world coordinate system. (Pitch and roll are assumed zero.)
velocity - 3D velocity vector components in meters per second (m/s) in the world coordinate system, representing the object's instantaneous motion along the x, y, and z axes respectively.

Our AI models are designed and/or optimized to run on NVIDIA GPU-accelerated systems. By leveraging NVIDIA’s hardware (e.g. GPU cores) and software frameworks (e.g., CUDA libraries), the model achieves faster training and inference times compared to CPU-only solutions.

Software Integration:

Runtime Engine(s):

DeepStream - 7.1
TAO - 6.1.0

Supported Hardware Microarchitecture Compatibility:

NVIDIA Blackwell
NVIDIA Hopper
NVIDIA Lovelace
NVIDIA Ampere

Preferred/Supported Operating Systems:

Linux

These models are designed for use with NVIDIA hardware and software. For hardware, the models are compatible with any NVIDIA DataCenter GPU. For software, the models are specifically designed for the TAO Toolkit.

Model versions (Latest):

trainable_v2.2 (ResNet50 / ResNet101) - Pre-trained model for Sparse4D.
deployable_v2.2 (ResNet50 / ResNet101) - Model for Sparse4D deployable to DeepStream or TensorRT.

Training, Testing, and Evaluation Datasets

Dataset Overview

** Total Size:

\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n

tr:nth-child(odd)]:bg-surface-sunken\">

Subset	No. of scenes	No. BEV groups	No. of objects per scene	No. of cameras per scene	Duration of each camera sequence	FPS
Train	90	3469	1-120	1-20	5 mins	30 FPS

** Total Number of Datasets: 3 Datasets in total including (MTMC Tracking 2025 and COSMOS augmented version of the MTMC Tracking 2025 dataset, SCOUT dataset)

** Dataset partition: Training 80%, Testing 10%, Validation 10%
\n** Time period for training data collection: Collected throughout the year 2024-2025.
\n** Time period for testing data collection Collected throughout the year 2024-2025.
\n** Time period for validation data collection Collected throughout the year 2024-2025.

The model is trained on the MTMC Tracking 2025 dataset available on HuggingFace.

Data Format & Preprocessing

Raw Data (MTMC Tracking 2025/AI City Challenge Format)

This raw data includes .mp4 RGB video files, depth maps stored in HDF5 format, ground_truth.json, calibration.json & map.png. For more details on these files including the ground truth format please refer to the raw dataset format here.

The raw data may also be obtained from Nvidia Isaac Sim Replicator Agent IRA. When collected via this route, RGB image data & depth maps can be in frame output/HDF5 output. Utilize the TAO Data Service to convert your raw data format (AICity) to pickle format (OVPKL) accordingly.

Pickle File Structure (OVPKL format)

The above raw data format needs to be converted to a pickle files for model training via the TAO DataService. The model will utilize these pickle files along with the above files for training. An example command can be found in the sparse4d finetuning notebook. The pickle file will consist of the items.

The process generates pickle files (.pkl) containing scene information. Each pickle file is a dictionary with the following top-level keys:

metadata: (dict) Contains metadata about the dataset.
infos: (list) A list of dictionaries, where each dictionary contains information about a specific frame in the scene.

`metadata` key

The metadata dictionary contains the following keys:

\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n

tr:nth-child(odd)]:bg-surface-sunken\">

Key	Type	Description	Example Value
`version`	`str`	Version of the dataset or split.	`'trainval_split'`
`split_type`	`str`	Type of data split (e.g., 'all', 'train', 'validation').	`'all'`

`infos` key

The infos key holds a list of dictionaries. Each dictionary in this list corresponds to a frame and contains the following keys:

\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n

tr:nth-child(odd)]:bg-surface-sunken\">

Key	Type	Description	Example Value (Illustrative)
`frame_idx`	`int`	Index of the frame.	`0`
`cams`	`dict`	Dictionary containing data for each camera in the frame. Keys are camera names (e.g., `'Camera'`, `'Camera_01'`).	See cams Structure below.
`scene_name`	`str`	Name of the scene.	`'Warehouse_014+bev-sensor-training-1'`
`timestamp`	`float`	Timestamp of the frame.	`0.0`
`token`	`str`	Unique token for the frame.	`'Warehouse_014+bev-sensor-training-1__000000000'`
`group_name`	`str`	Name of the group this frame belongs to.	`'bev-sensor-training-1'`
`instance_inds`	`np.ndarray`	Array of instance indices present in the frame.	`array([372, 171, 172, 631, 663])`
`asset_inds`	`np.ndarray`	Array of asset indices corresponding to instances.	`array([ 771, 772, 775, 2094, 1372])`
`gt_boxes`	`np.ndarray`	Ground truth bounding boxes for objects in the frame. Shape: (N, 7) where N is the number of objects. Each row is `[x, y, z, dx, dy, dz, heading]`.	`array([[-0.686, 1.073, ..., -0. ], ...])`
`gt_names`	`np.ndarray`	Array of ground truth names for objects.	`array(['agility_digit', 'gr1_t2', ...], dtype='<U18')`
`gt_velocity`	`np.ndarray`	Ground truth velocity for objects. Shape: (N, 3) for `[vx, vy, vz]`.	`array([[0., 0., 0.], ...])`
`valid_flag`	`np.ndarray`	Boolean array indicating if the ground truth data for each object is valid.	`array([ True, True, ...])`
`gt_visibility`	`list`	List of dictionaries, one for each ground truth object, indicating its visibility percentage in each camera.	`[{'Camera': 1.0, 'Camera_01': 1.0, ...}, ...]`

`cams` Structure

The cams dictionary (within each element of the infos list) contains nested dictionaries, where each key is a camera identifier (e.g., 'Camera', 'Camera_01', etc.). Each of these camera-specific dictionaries has the following structure:

\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n

tr:nth-child(odd)]:bg-surface-sunken\">

Key	Type	Description	Example Value (Illustrative)
`data_path`	`tuple`	Tuple containing paths to camera data. `(h5_file_path, rgb_image_relative_path)`	`('data/mtmc/Warehouse_014/Camera.h5', 'rgb/rgb_00000.jpg')`
`depth_map_path`	`tuple`	Tuple containing paths to depth map data. `(h5_file_path, depth_image_relative_path)`	`('data/mtmc/Warehouse_014/Camera.h5', 'distance_to_image_plane_png/distance_to_image_plane_00000.png')`
`sample_data_token`	`str`	Unique token for the sample data from this camera.	`'Warehouse_014+bev-sensor-training-1__000000000+Camera'`
`cam_intrinsic`	`np.ndarray`	3x3 camera intrinsic matrix.	`array([[916.249, 0., 960.], [0., 916.249, 540.], [0., 0., 1.]])`
`sensor2world_transform`	`np.ndarray`	4x4 transformation matrix from sensor coordinates to world coordinates.	`array([[0.018, -0.999, ..., -4.731], ..., [0., 0., 0., 1.]])`

3D bounding box information is transformed from OV coordinates to BEV coordiantes for faster convergence during training. During inference, results are converted back to OV world coordinates.

Public Datasets

MTMC Tracking 2025 Subset from NVIDIA PhysicalAI-SmartSpaces dataset

Private Datasets

NVIDIA COSMOS-augmented version of the MTMC Tracking 2025 dataset was used for training to improve data diversity and robustness across varying scenarios.

Self-Sourced Synthetic Data

Overall Size per Modality: 80% of the overall data used for training was synthetic data collected from NVIDIA Isaac Sim Replicator.
A description of the methods used to synthetically generate training data: Various warehouse related assets were imported in a scene with virtual cameras recording image data.

Training Data

The dataset statistics used for pretraining the model are as follows:

\n
Data Modality - Image
\n
\n
Image Training Data Size - 1 Million to 1 Billion Images
\n
\n
Data Collection Method by dataset - Synthetic
\n
\n
Labeling Method by dataset - Synthetic
\n

Properties (Quantity, Dataset Descriptions, Sensor(s)):

Please see above.

Model was trained on the following moving object classes: Person, Fourier_GR1_T2_Humanoid, Agility_Digit_Humanoid, Nova_Carter, Transporter and Forklift.
Both validation & testing was conduction on random scenes from the MTMC Tracking 2025 subset.

Testing and Evaluation Datasets

Both validation & testing was conduction on random scenes from the MTMC Tracking 2025 subset.

\n
Data Collection Method by dataset - Synthetic
\n
\n
Labeling Method by dataset - Synthetic
\n

Properties (Quantity, Dataset Descriptions, Sensor(s)):\nWe utilize a random scenes from the test set of the MTMC Tracking 2025 subset.

Methodology and KPI

The key performance indicator is Higher Order Tracking Accuracy (HOTA) per-class and the average HOTA obtained across all classes. We utilize the HOTA evaluation methodology to assess multi-object tracking accuracy.

Higher Order Tracking Accuracy (HOTA) is a metric that balances detection accuracy and association accuracy into a single unified score. It decomposes tracking performance into Detection Accuracy (DetA), which measures how well objects are localized, and Association Accuracy (AssA), which measures how well detections are linked over time into consistent tracks. HOTA is computed as the geometric mean of DetA and AssA, averaged over a range of localization thresholds. This provides a balanced, holistic measure of tracking quality that captures both spatial accuracy and temporal consistency.

The following scores are for v2.2 models trained on MTMC Tracking 2025 subset. The evaluation set and training set is disjoint.

ResNet50 backbone (v2.2)

\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n

tr:nth-child(odd)]:bg-surface-sunken\">

Object Class	HOTA	DetA	AssA	LocA
`Person`	53.93	59.13	49.54	73.34
`Fourier_GR1_T2_Humanoid`	59.31	59.87	58.84	73.46
`Agility_Digit_Humanoid`	52.69	52.57	53.50	70.72
`Nova_Carter`	58.26	56.89	59.92	71.05
`Transporter`	33.70	38.98	31.49	68.41
`Forklift`	48.05	43.84	54.97	66.69
Average	50.99	51.88	51.37	70.61

ResNet101 backbone (v2.2)

\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n

tr:nth-child(odd)]:bg-surface-sunken\">

Object Class	HOTA	DetA	AssA	LocA
`Person`	57.96	58.99	57.11	73.29
`Fourier_GR1_T2_Humanoid`	60.93	60.97	60.98	74.00
`Agility_Digit_Humanoid`	54.00	53.32	55.32	71.02
`Nova_Carter`	54.85	57.60	53.28	70.92
`Transporter`	33.98	38.13	32.58	67.33
`Forklift`	48.05	43.78	55.20	66.45
Average	51.63	52.13	52.41	70.50

Real-time Inference Performance

Inference runs through the DeepStream pipeline on TensorRT with mixed precision (FP16+FP32). The TensorRT columns capture model-only latency, while the DeepStream columns add the instance-bank pre- and post-processing overhead. The table summarizes how many cameras each GPU supports at 30, 15, and 10 FPS for the v2.0 model. Numbers in bold are measured (all TensorRT values and certain DeepStream values). DeepStream values not in bold are estimates based on the guidance that the DS microservice adds approximately 30% overhead over Sparse4D TensorRT model performance.

\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n

tr:nth-child(odd)]:bg-surface-sunken\">

GPU	TensorRT @30 FPS	DeepStream @30 FPS	TensorRT @15 FPS	DeepStream @15 FPS	TensorRT @10 FPS	DeepStream @10 FPS
1 x A100-SXM4-80GB	13	9	28	19	43	30
1 x B200	75	52	155	108	200	140
1 x GB200	101	70	200	140	200	140
1 x H100 NVL - 94GB	23	16	48	33	72	50
1 x H100 SXM HBM3 - 80GB	28	19	58	40	89	62
1 x H200	32	22	67	46	103	72
1 x L4 - 24GB	3	2	6	4	9	6
1 x L40	8	5	16	11	26	18
1 x L40S - 48GB	9	6	20	14	30	21
1 x RTX 6000 ADA	9	6	19	13	27	18
1 x RTX PRO 6000 Blackwell (Server)	28	19	58	40	89	62
1 x RTX PRO 6000 Blackwell (Workstation)	29	20	59	41	90	63
1 x IGX Thor - T7000 iGPU (no dGPU)	5	3	11	7	17	11
1 x IGX Thor - T7000 dGPU - RTX PRO 6000 Blackwell Max-Q Workstation Edition	21	14	30	21	30	21
1 x AGX Thor - T5000	4	2	8	5	14	9
1 x DGX Spark	5	3	11	7	17	11

Changelog

v2.2

Released two backbone variants of the same recipe: ResNet50 (lower latency / smaller footprint) and ResNet101 (high accuracy), shipped under sparse4d_rn50 and sparse4d_rn101 NGC model cards respectively.
The ResNet50 variant is built on top of an RT-DETR R50 pre-training initialization for stronger image features at half the backbone cost.
Added pallet_truck to the trained class set (person, gr1_t2, agility_digit, nova_carter, transporter, forklift, pallet_truck).
Expanded the training corpus to all available MTMC Tracking 2025 scenes (lazy-loaded), 3K+ BEV groups, and refreshed COSMOS-augmented data; mixed FPS training is unchanged.
Refreshed kmeans anchor priors per backbone (_ov_kmeans900_v2.2_r50.npy, _ov_kmeans900_v2.2_r101.npy).

v2.1

Trained on a significantly expanded dataset: more synthetic environments (retail, hospital, supermarket), GTC'24 warehouse data, COSMOS-transferred data with BEV groups, and real-world SCOUT data.
Leverages the upgraded camera grouping algorithm (SDU v1.2).
Corrected forklift orientation annotations.

v2.0

Increased the number of BEV groups to 1337 to improve accuracy across all benchmarks.
Added COSMOS-augmented datasets to the training set, using an 80:20 split between synthetic data and COSMOS augmentations.
Introduced multiple FPS settings in training to better handle low-frame-rate streams (1, 2, 3, 5, 6, 10, 15, and 30 FPS).
Increased the regression weight for rotation loss to improve bounding box fitting accuracy.

v1.1

Updated input image resolution to 960×540 and introduced camera embeddings to lift maximum camera limits.
Expanded training set to 34 scenes spanning 500+ cameras and 380 BEV groups (1–18 cameras each).
Supported classes: Person, Fourier_GR1_T2_Humanoid, Agility_Digit_Humanoid, Nova_Carter, Transporter, Forklift.
Mixed input frame rates (10, 15, 30 FPS) to improve robustness to lower streaming FPS.
Enabled sparser camera grouping with random frame dropping to mirror Data Service frame-loss scenarios.
Fixed rotation bug in object annotations and retrained the model with corrected labels.
New MSDA FP16 Kernel which helps improve throughput by ~1.5x-2x for Data Center GPUs available in the DeepStream Pipeline inside the Nvidia Video Search and Summarization (VSS) - Warehouse Blueprint release.

v1.0

First commercial warehouse-focused 3D detection and tracking release.
Input image resolution: 1408×512 and cameras limited to 20.
Supported classes: Person, Fourier_GR1_T2_Humanoid, Agility_Digit_Humanoid, Nova_Carter.

How to use this model

In order to use the model as pre-trained weights for transfer learning, please use the snippet below as a template for the model component of the experiment spec file to train a Sparse4D. For more information on experiment spec file, please refer to the Train Adapt Optimize (TAO) Toolkit User Guide.

model:\n  type: \"sparse4d\"\n  use_grid_mask: true\n  use_deformable_func: true\n  use_temporal_align: true\n  input_shape: [960, 540]\n  embed_dims: 256\n  neck:\n    type: \"FPN\"\n    num_outs: 4\n    start_level: 0\n    out_channels: 256\n    in_channels: [256, 512, 1024, 2048]\n    add_extra_convs: \"on_output\"\n    relu_before_extra_convs: true\n  depth_branch:\n    type: \"dense_depth\"\n    embed_dims: \"${model.embed_dims}\"\n    num_depth_layers: 3\n    loss_weight: 0.2\n  head:\n    type: \"sparse4d\"\n    num_output: 300\n    cls_threshold_to_reg: 0.05\n    decouple_attn: true\n    return_feature: true\n    use_reid_sampling: false\n    embed_dims: \"${model.embed_dims}\"\n    num_groups: 8\n    num_decoder: 6\n    num_single_frame_decoder: 1\n    drop_out: 0.1\n    temporal: true\n    with_quality_estimation: true\n    instance_bank:\n      num_anchor: 900\n      anchor: ???\n      num_temp_instances: 600\n      confidence_decay: 0.8\n      feat_grad: false\n      default_time_interval: 0.033333\n      embed_dims: \"${model.embed_dims}\"\n      use_temporal_align: \"${model.use_temporal_align}\"\n    anchor_encoder:\n      type: 'SparseBox3DEncoder'\n      vel_dims: 3\n      embed_dims: [128, 32, 32, 64]\n      mode: 'cat'\n      output_fc: false\n      in_loops: 1\n      out_loops: 4\n    operation_order: [\n      \"deformable\", \"ffn\", \"norm\", \"refine\", \"temp_gnn\", \"gnn\", \"norm\", \n      \"deformable\", \"ffn\", \"norm\", \"refine\", \"temp_gnn\", \"gnn\", \"norm\", \n      \"deformable\", \"ffn\", \"norm\", \"refine\", \"temp_gnn\", \"gnn\", \"norm\", \n      \"deformable\", \"ffn\", \"norm\", \"refine\", \"temp_gnn\", \"gnn\", \"norm\", \n      \"deformable\", \"ffn\", \"norm\", \"refine\", \"temp_gnn\", \"gnn\", \"norm\", \n      \"deformable\", \"ffn\", \"norm\", \"refine\"\n    ]\n    temp_graph_model:\n      type: \"MultiheadAttention\"\n      embed_dims: 512\n      num_heads: 8\n      batch_first: true\n      dropout: 0.1\n    graph_model:\n      type: \"MultiheadAttention\"\n      embed_dims: \"${model.head.temp_graph_model.embed_dims}\"\n      num_heads: \"${model.head.temp_graph_model.num_heads}\"\n      batch_first: true\n      dropout: \"${model.head.temp_graph_model.dropout}\"\n    norm_layer:\n      type: \"LN\"\n      normalized_shape: \"${model.embed_dims}\"\n    ffn:\n      type: \"AsymmetricFFN\"\n      in_channels: 512\n      pre_norm:\n        type: \"LN\"\n      embed_dims: 256\n      feedforward_channels: 1024\n      num_fcs: 2\n      ffn_drop: 0.1\n      act_cfg:\n        type: \"ReLU\"\n        inplace: true\n    deformable_model:\n      embed_dims: \"${model.embed_dims}\"\n      num_groups: 8\n      num_levels: 4\n      attn_drop: 0.15\n      use_deformable_func: true\n      use_camera_embed: true\n      residual_mode: \"cat\"\n      kps_generator:\n        embed_dims: \"${model.embed_dims}\"\n        num_learnable_pts: 6\n        fix_scale:\n          - [0, 0, 0]\n          - [0.45, 0, 0]\n          - [-0.45, 0, 0]\n          - [0, 0.45, 0]\n          - [0, -0.45, 0]\n          - [0, 0, 0.45]\n          - [0, 0, -0.45]\n    refine_layer:\n      type: \"SparseBox3DRefinementModule\"\n      embed_dims: \"${model.embed_dims}\"\n      refine_yaw: true\n      with_quality_estimation: true\n    sampler:\n      num_dn_groups: 5\n      num_temp_dn_groups: 3\n      dn_noise_scale: [2.0, 2.0, 2.0, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5]\n      max_dn_gt: 128\n      add_neg_dn: true\n      cls_weight: 2.0\n      box_weight: 0.25\n      reg_weights: [2.0, 2.0, 2.0, 0.5, 0.5, 0.5, 0.0, 0.0, 0.0, 0.0, 0.0]\n      use_temporal_align: \"${model.use_temporal_align}\"\n    visibility_net:\n      type: \"visibility_net\"\n      embedding_dim: 256\n      hidden_channels: 32\n    loss:\n      reg:\n        type: \"sparse_box_3d\"\n        box_weight: 0.25\n        cls_allow_reverse: [5, 6, 7]\n      cls:\n        type: \"focal\"\n        use_sigmoid: true\n        gamma: 2.0\n        alpha: 0.25\n        loss_weight: 2.0\n      id:\n        type: \"cross_entropy_label_smooth\"\n        num_ids: \"${dataset.num_ids}\"\n    bnneck:\n      type: \"bnneck\"\n      feat_dim: 256\n      num_ids: \"${dataset.num_ids}\"\n    decoder:\n      type: \"SparseBox3DDecoder\"\n      score_threshold: 0.05\n    reg_weights: [2.0, 2.0, 2.0, 1 ,1, 1, 1, 1, 1, 1, 1]\n\n

Inference & Deployment:

For deployment of this model, please refer to the our Nvidia Video Search and Summarization (VSS) - Warehouse Blueprint release documentation.

Acceleration Engine: Tensor(RT)

Test Hardware:

NVIDIA Datacenter GPUs

Ethical Considerations

NVIDIA believes Trustworthy AI is a shared responsibility and we have established policies and practices to enable development for a wide array of AI applications. When downloaded or used in accordance with our terms of service, developers should work with their internal model team to ensure this model meets requirements for the relevant industry and use case and addresses unforeseen product misuse.

Please make sure you have proper rights and permissions for all input image and video content; if image or video includes people, personal health information, or intellectual property, the image or video generated will not blur or maintain proportions of image subjects included.

For more detailed information on ethical considerations for this model, please see the Model Card++ Explainability, Bias, Safety & Security, and Privacy Sub-cards.

Please report model quality, risk, security vulnerabilities or NVIDIA AI Concerns here.

","resourceVersions",[206,207,208,209,210,211,212,213,214,215,216],"resourceVersionsTotal",10,"hasMoreVersions","versionMetadata",{"_179":22,"_180":181,"_49":182,"_183":-5,"_184":185,"_186":187,"_188":189,"_190":16,"_191":98,"_192":-5,"_193":31,"_194":-5,"_195":31,"_196":-5,"_197":31,"_198":-5,"_199":31,"_200":-5,"_201":31,"_202":-5,"_203":31,"_204":205},"versionNotFound","isLoggedIn","latestVersionPath","/orgs/nvidia/tao/models/sparse4d_rn101/-","version","imageReference","cliDownloadCommand","ngc registry model download-version nvidia/tao/sparse4d_rn101:trainable_v2.0","helmFetchCommand","viewAllVersionsPath","/orgs/nvidia/tao/models/sparse4d_rn101/trainable_v2.0/version-history","accessCheckResult",{"_178":98},"hasRelatedCollections","context",{"_174":175,"_176":177},1787683775201,[],[],[167,168,169,170,171,172,173],{"_45":59,"_47":70,"_49":61,"_51":62,"_53":71,"_54":55},{"_45":59,"_47":68,"_49":61,"_51":62,"_53":69,"_54":55},{"_45":59,"_47":66,"_49":61,"_51":62,"_53":67,"_54":55},{"_45":59,"_47":64,"_49":114,"_51":-5,"_53":-5,"_54":55},{"_45":59,"_47":60,"_49":61,"_51":62,"_53":63,"_54":55},{"_45":56,"_47":57,"_49":50,"_51":52,"_53":58,"_54":55},{"_45":46,"_47":48,"_49":50,"_51":52,"_53":16,"_54":55},"orgName","nvidia","teamName","tao","allowed","versionId","description","","UPLOAD_COMPLETE","releaseType","createdDate",["D",1770617866216],"updatedDate",["D",1770617873449],"totalSizeInBytes",804541564,"totalFileCount","isSigned","malwareScanStatus","isScanned","releaseNotes","hasReleaseNotes","setup","hasSetup","quickStartGuide","hasQuickStartGuide","advanced","hasAdvanced","performance","hasPerformance","policies",[],{"_78":236,"_80":237,"_183":-5,"_191":98,"_193":31,"_204":233,"_195":31},{"_78":234,"_80":234,"_183":-5,"_191":98,"_193":31,"_204":235,"_195":31},{"_78":232,"_80":232,"_183":-5,"_191":98,"_193":31,"_204":233,"_195":31},{"_78":230,"_80":230,"_183":-5,"_191":98,"_193":31,"_204":231,"_195":31},{"_78":228,"_80":228,"_183":-5,"_191":98,"_193":31,"_204":229,"_195":31},{"_78":226,"_80":226,"_183":-5,"_191":98,"_193":31,"_204":227,"_195":31},{"_78":22,"_80":22,"_183":-5,"_191":98,"_193":31,"_204":225,"_195":31},{"_78":223,"_80":223,"_183":-5,"_191":98,"_193":31,"_204":224,"_195":31},{"_78":221,"_80":221,"_183":-5,"_191":98,"_193":31,"_204":222,"_195":31},{"_78":219,"_80":219,"_183":-5,"_191":98,"_193":31,"_204":220,"_195":31},{"_78":217,"_80":217,"_183":-5,"_191":98,"_193":31,"_204":218,"_195":31},"deployable_v1.0",[],"trainable_v1.0",[],"trainable_v1.1",[],"deployable_v1.1",[],[],"deployable_v2.0",[],"trainable_v2.1",[],"deployable_v2.1",[],"trainable_v2.2",[],"deployable_v2.2",[],"-","Latest (trainable_v2.2)","nvidia/tao/sparse4d_rn101","key","nvidia/tao/models/sparse4d_rn101","resourceType","models","resourceId","sparse4d_rn101","shortDescription","TAO Pretrained Sparse4D with ResNet-101 Backbone","# TAO Sparse4D Model Card\n\n## Model Overview \n\n### Description:\nTAO Sparse4D is an advanced three-dimensional (3D) multi-camera detection and tracking network. It generates precise 3D bounding boxes and tracking IDs for a diverse set of objects across multiple camera views. The model was developed by NVIDIA as a part of the TAO model family. The provided model is pre-trained on the Multi-Target Multi-Camera (MTMC) Tracking 2025 subset from the [NVIDIA PhysicalAI-SmartSpaces dataset](https://huggingface.co/datasets/nvidia/PhysicalAI-SmartSpaces/tree/main) also utilized for the 2025 AI City Challenge. This model is ready for commercial use.\n\n\n

\n\nThe model in this card was trained & evaluated on the following warehouse-centric classes: `Person`, `Fourier_GR1_T2_Humanoid`, `Agility_Digit_Humanoid`, `Nova_Carter`, `Transporter` and `Forklift`.\n\n### License/Terms of Use: \n\nGOVERNING TERMS: Use of this model is governed by the [NVIDIA Open Model License](https://www.nvidia.com/en-us/agreements/enterprise-software/nvidia-open-model-license/).\n\n### Deployment Geography:\nGlobal\n
\n\n### Use Case:
\nTAO Sparse4D is designed for 3D multi-camera object detection and tracking in indoor environments like warehouses and logistics facilities. The model detects & tracks objects across multiple camera views for applications including warehouse automation, safety and workflow optimization in industrial settings, providing spatial understanding of a scene.\n\n\n### Release Date:
\nv1.0 Model - NGC - 06/13/2025
\nv1.1 Model - NGC - 11/30/2025
\nv2.0 Model - NGC - 01/23/2026 via [NGC](https://catalog.ngc.nvidia.com/orgs/nvidia/teams/tao/models/sparse4d_rn101/files?version=deployable_v2.0)
\nv2.1 Model - NGC - 02/27/2026 via [NGC](https://catalog.ngc.nvidia.com/orgs/nvidia/teams/tao/models/sparse4d_rn101/files?version=deployable_v2.1)
\nv2.2 Model (ResNet50) - NGC - 05/12/2026 via [NGC](https://catalog.ngc.nvidia.com/orgs/nvidia/teams/tao/models/sparse4d_rn50/files?version=deployable_v2.2)
\nv2.2 Model (ResNet101) - NGC - 05/12/2026 via [NGC](https://catalog.ngc.nvidia.com/orgs/nvidia/teams/tao/models/sparse4d_rn101/files?version=deployable_v2.2)
\n\n\n## References \n\n- [Wang, Y., Pusegaonkar, S., Wang, Y., Li, A., Kumar, V., Sethi, C., ... & Biswas, S. (2026). A Unified 3D Object Perception Framework for Real-Time Outside-In Multi-Camera Systems. arXiv preprint arXiv:2601.10819.](https://arxiv.org/abs/2601.10819)\n- [Lin, X., Pei, Z., Lin, T., Huang, L., & Su, Z. (2023). Sparse4d v3: Advancing end-to-end 3d detection and tracking. arXiv preprint arXiv:2311.11722.](https://arxiv.org/abs/2311.11722)\n\n\n## Model Architecture \n\n**Architecture Type:** Convolution Neural Network (CNN) based backbone with Transformer based decoder layers.
\n\n**Network Architecture:** ResNet101 backbone with Transformer based decoder layers.
\n\n**Number of model parameters:** 7.8 × 10⁷ parameters\n\nThis work leverages the [Sparse4D v3 model](https://arxiv.org/abs/2311.11722) & is tailored for indoor environments like warehouses with static camera setups. Sparse4D is a query based technique which involves sampling sparse features for better computational efficiency compared to other dense 3D detection & tracking based techniques. The architecture features a ResNet101 backbone and processes time-synchronized frames from multiple cameras. Key components include the backbone, a Feature Pyramid Network (FPN), and multiple decoder layers incorporating Multi-Scale Deformable Aggregation blocks (featuring key-point generation, feature sampling, visibility net module & embedding generation) along with refinement & classification layers. The model is trained on regression, classification & ID losses. The model also consists of an Instance Bank module utilized for tracking ID assignment & management. Along with the 3D bounding box information & object ID, Sparse4D also outputs an instance feature containing high-dimensional semantics from the image encoder. The training algorithm optimizes the network to minimize the classification loss, regression loss & ID loss.\n\n

\n\n\n## Input:
\n\nSparse4D is trained on RGB images, camera calibration files, ground truth 3D bounding boxes & object ids, and optional depth maps. Since performing 3D multi-camera detection and tracking across large regions like warehouses with high-density camera setups is computationally expensive, we partition large regions into overlapping groups called Bird's Eye View (BEV) group. Each BEV group contains multiple cameras and serves as the fundamental training unit for Sparse4D. \n\n**Input Type(s):** Each BEV group will have cameras with RGB images, Camera calibration file, Ground Truth with 3D bounding boxes & Optional Depth Maps available via HuggingFace or generated synthetically by Nvidia Isaac Sim Replicator.
\n\n**Input Format:** Red, Green, Blue (RGB) images stored in raw png/jpg/hdf5 and Depth Maps stored in png/hdf5. Can support input resolution of 3 x 1080 x 1920 for both RGB & Depth Maps. Data preprocessing is required for grouping images & depth maps from multiple cameras to appropriate BEV groups. This can be done via the TAO Data Service.
\n\n**Input Parameters:** Multiple dimensions. See below for detailed model input shapes.
\n\n**Other Properties Related to Input:** 3 x 1080 x 1920 (C x H x W) resolution images for both RGB & Depth Maps. Data Pre-Processing needed via TAO Data Services. No alpha channel required.
\n\nNote that depth maps are optional and not required for evaluation or inference. The raw model inputs are as follows:\n\n| Dimension | Description |\n|-----------|-------------|\n| `B` | Batch size |\n| `C` | Number of channels |\n| `N` | Number of cameras |\n| `H` | Image height |\n| `W` | Image width |\n| `Q` | Number of queries |\n| `M` | Number of output boxes |\n| `E` | Number of instance features |\n\n| Input Name | Type | Shape | ResNet101 Shape | Description |\n|------------|------|-------|---------------|-------------|\n| `img` | List[Tensor] | (B, N, C, H, W) | (1, N, 3, 540, 960) | Input image tensor |\n| `projection_mat` | List[Tensor] | (B, N, 4, 4) | (1, 10, 4, 4) | List of projection matrices |\n| `image_wh` | List[Tensor] | (B, N, 2) | (1, 10, 2) | List of image width and height |\n| `input_cached_feature` | List[Tensor] | (B, 600, 256) | (1, 600, 256) | List of cached features |\n| `input_cached_anchor` | List[Tensor] | (B, 600, 11) | (1, 600, 11) | List of cached anchor |\n| `prev_exists` | List[Tensor] | (B) | (1) | Indicates if previous frame exists or not |\n| `interval_mask` | List[Tensor] | (Bx1x1) | (1x1x1) | Boolean to describe the interval mask |\n\n\n* No. of cameras is dynamic in the model.\n* Model is initialized with zero tensors for the `input_cached_feature`, `input_cached_anchor`, `prev_exists` & `interval_mask` at the first frame. These tensors will be updated automatically via instance bank from the second frame & onwards.\n\n

\n\n\n## Output:
\n**Output Type(s):** Tensors consisting of bounding boxes in 3D , object confidence scores, classes, class confidence scores, tracking object ID & instance features.
\n\n**Output Format:** List of Tensors.
\n\n**Output Parameters:** Multiple dimensions. See below for detailed model output shapes.
\n\n**Other Properties Related to Output:** Please see the details below.
\n\nThe final output is a list with length of batch size after post-processing. Each element is a dictionary with the following keys:\n\n| Output Name | Type | Shape | ResNet Shape | Description |\n|------------|------|-------|---------------|-------------|\n| `boxes_3d` | List[Tensor] | (B, M, 10) | (1, 600, 10) | List of 3D boxes: `(x, y, z, width, length, height, yaw, velocity_x, velocity_y, velocity_z)` in OV coordinates |\n| `scores_3d` | List[Tensor] | (B, M) | (1, 600) | List of object confidence scores (classification scores * centerness score) |\n| `labels_3d` | List[Tensor] | (B, M) | (1, 600) | List of class labels |\n| `cls_scores` | List[Tensor] | (B, M) | (1, 600) | List of classification scores |\n| `instance_ids` | List[Tensor] | (B, M) | (1, 600) | List of instance IDs |\n| `instance_feats` | List[Tensor] | (B, M, E) | (1, 600, 256) | List of instance features |\n\nwhere:\n\n- **x, y, z** - 3D coordinates of the bounding-box centroid in the world coordinate system which is in meters.\n- **width, length, height** - Box dimensions in meters along its x (width), y (length) and z (height) axes of the object-centered coordinate system, with the origin at the centroid.\n- **yaw** - Euler angle in radians about the y-axis of the object-centered coordinate system defining the box’s heading in the world coordinate system. (Pitch and roll are assumed zero.)\n- **velocity** - 3D velocity vector components in meters per second (m/s) in the world coordinate system, representing the object's instantaneous motion along the x, y, and z axes respectively.\n\n

\n\n\nOur AI models are designed and/or optimized to run on NVIDIA GPU-accelerated systems. By leveraging NVIDIA’s hardware (e.g. GPU cores) and software frameworks (e.g., CUDA libraries), the model achieves faster training and inference times compared to CPU-only solutions. \n\n## Software Integration:\n\n**Runtime Engine(s):** \n* DeepStream - 7.1
\n* TAO - 6.1.0
\n\n**Supported Hardware Microarchitecture Compatibility:**
\n* NVIDIA Blackwell
\n* NVIDIA Hopper
\n* NVIDIA Lovelace
\n* NVIDIA Ampere
\n\n**Preferred/Supported Operating Systems:**
\n* Linux
\n\nThese models are designed for use with NVIDIA hardware and software. For hardware, the models are compatible with any NVIDIA DataCenter GPU. For software, the models are specifically designed for the [TAO Toolkit](https://github.com/NVIDIA/tao_pytorch_backend).\n\n## Model versions (Latest):\n\n- **trainable_v2.2** (ResNet50 / ResNet101) - Pre-trained model for Sparse4D.\n- **deployable_v2.2** (ResNet50 / ResNet101) - Model for Sparse4D deployable to DeepStream or TensorRT.\n\n\n## Training, Testing, and Evaluation Datasets \n\n### Dataset Overview\n** Total Size:\n\n|Subset|No. of scenes|No. BEV groups|No. of objects per scene|No. of cameras per scene|Duration of each camera sequence|FPS|\n|:---:|:---:|:---:|:---:|:---:|:---:|:---:|\n|Train|90|3469|1-120|1-20|5 mins|30 FPS|\n\n** Total Number of Datasets: 3 Datasets in total including (MTMC Tracking 2025 and COSMOS augmented version of the MTMC Tracking 2025 dataset, SCOUT dataset)\n\n** Dataset partition: Training 80%, Testing 10%, Validation 10%
\n** Time period for training data collection: Collected throughout the year 2024-2025.
\n** Time period for testing data collection Collected throughout the year 2024-2025.
\n** Time period for validation data collection Collected throughout the year 2024-2025.
\n\nThe model is trained on the MTMC Tracking 2025 dataset available on [HuggingFace](https://huggingface.co/datasets/nvidia/PhysicalAI-SmartSpaces/tree/main/MTMC_Tracking_2025). \n\n### Data Format & Preprocessing\n\n#### Raw Data (MTMC Tracking 2025/AI City Challenge Format)\n\nThis raw data includes .mp4 RGB video files, depth maps stored in HDF5 format, ground_truth.json, calibration.json & map.png. For more details on these files including the ground truth format please refer to the raw dataset format [here](https://huggingface.co/datasets/nvidia/PhysicalAI-SmartSpaces#directory-structure-for-mtmc_tracking_2025). \n\nThe raw data may also be obtained from Nvidia Isaac Sim Replicator Agent [IRA](https://docs.isaacsim.omniverse.nvidia.com/latest/replicator_tutorials/tutorial_replicator_agent.html). When collected via this route, RGB image data & depth maps can be in frame output/HDF5 output. Utilize the TAO Data Service to convert your raw data format (AICity) to pickle format (OVPKL) accordingly.\n\n#### Pickle File Structure (OVPKL format)\n\nThe above raw data format needs to be converted to a pickle files for model training via the TAO DataService. The model will utilize these pickle files along with the above files for training. An example command can be found in the sparse4d finetuning notebook. The pickle file will consist of the items.\n\n\nThe process generates pickle files (`.pkl`) containing scene information. Each pickle file is a dictionary with the following top-level keys:\n\n- `metadata`: (dict) Contains metadata about the dataset.\n- `infos`: (list) A list of dictionaries, where each dictionary contains information about a specific frame in the scene.\n\n##### `metadata` key\n\nThe `metadata` dictionary contains the following keys:\n\n| Key | Type | Description | Example Value |\n| :--------- | :----- | :------------------------------------------- | :------------------------------------ |\n| `version` | `str` | Version of the dataset or split. | `'trainval_split'` |\n| `split_type` | `str` | Type of data split (e.g., 'all', 'train', 'validation'). | `'all'` |\n\n##### `infos` key\n\nThe `infos` key holds a list of dictionaries. Each dictionary in this list corresponds to a frame and contains the following keys:\n\n| Key | Type | Description | Example Value (Illustrative) |\n| :---------------- | :---------- | :--------------------------------------------------------------------------------------------------------- | :--------------------------------------------------------------- |\n| `frame_idx` | `int` | Index of the frame. | `0` |\n| `cams` | `dict` | Dictionary containing data for each camera in the frame. Keys are camera names (e.g., `'Camera'`, `'Camera_01'`). | See [cams Structure](#cams-structure) below. |\n| `scene_name` | `str` | Name of the scene. | `'Warehouse_014+bev-sensor-training-1'` |\n| `timestamp` | `float` | Timestamp of the frame. | `0.0` |\n| `token` | `str` | Unique token for the frame. | `'Warehouse_014+bev-sensor-training-1__000000000'` |\n| `group_name` | `str` | Name of the group this frame belongs to. | `'bev-sensor-training-1'` |\n| `instance_inds` | `np.ndarray`| Array of instance indices present in the frame. | `array([372, 171, 172, 631, 663])` |\n| `asset_inds` | `np.ndarray`| Array of asset indices corresponding to instances. | `array([ 771, 772, 775, 2094, 1372])` |\n| `gt_boxes` | `np.ndarray`| Ground truth bounding boxes for objects in the frame. Shape: (N, 7) where N is the number of objects. Each row is `[x, y, z, dx, dy, dz, heading]`. | `array([[-0.686, 1.073, ..., -0. ], ...])` |\n| `gt_names` | `np.ndarray`| Array of ground truth names for objects. | `array(['agility_digit', 'gr1_t2', ...], dtype='\n\n\n### Training Data \n\nThe dataset statistics used for pretraining the model are as follows:\n\n* Data Modality - Image
\n\n* Image Training Data Size - 1 Million to 1 Billion Images
\n\n* Data Collection Method by dataset - Synthetic
\n\n* Labeling Method by dataset - Synthetic
\n\n**Properties (Quantity, Dataset Descriptions, Sensor(s)):**\n\nPlease see above.\n\n* Model was trained on the following moving object classes: `Person`, `Fourier_GR1_T2_Humanoid`, `Agility_Digit_Humanoid`, `Nova_Carter`, `Transporter` and `Forklift`.\n* Both validation & testing was conduction on random scenes from the MTMC Tracking 2025 subset.\n\n\n## Testing and Evaluation Datasets\n\nBoth validation & testing was conduction on random scenes from the MTMC Tracking 2025 subset. \n\n* Data Collection Method by dataset - Synthetic
\n\n* Labeling Method by dataset - Synthetic
\n\n**Properties (Quantity, Dataset Descriptions, Sensor(s)):**\nWe utilize a random scenes from the test set of the MTMC Tracking 2025 subset.\n\n\n### Methodology and KPI\n\nThe key performance indicator is Higher Order Tracking Accuracy (HOTA) per-class and the average HOTA obtained across all classes. We utilize the HOTA evaluation methodology to assess multi-object tracking accuracy.\n\nHigher Order Tracking Accuracy (HOTA) is a metric that balances detection accuracy and association accuracy into a single unified score. It decomposes tracking performance into Detection Accuracy (DetA), which measures how well objects are localized, and Association Accuracy (AssA), which measures how well detections are linked over time into consistent tracks. HOTA is computed as the geometric mean of DetA and AssA, averaged over a range of localization thresholds. This provides a balanced, holistic measure of tracking quality that captures both spatial accuracy and temporal consistency.\n\nThe following scores are for v2.2 models trained on MTMC Tracking 2025 subset. The evaluation set and training set is disjoint.\n\n#### ResNet50 backbone (v2.2)\n\n| Object Class | HOTA | DetA | AssA | LocA |\n|---------------------------|-------|-------|-------|-------|\n| `Person` | 53.93 | 59.13 | 49.54 | 73.34 |\n| `Fourier_GR1_T2_Humanoid` | 59.31 | 59.87 | 58.84 | 73.46 |\n| `Agility_Digit_Humanoid` | 52.69 | 52.57 | 53.50 | 70.72 |\n| `Nova_Carter` | 58.26 | 56.89 | 59.92 | 71.05 |\n| `Transporter` | 33.70 | 38.98 | 31.49 | 68.41 |\n| `Forklift` | 48.05 | 43.84 | 54.97 | 66.69 |\n| Average | 50.99 | 51.88 | 51.37 | 70.61 |\n\n#### ResNet101 backbone (v2.2)\n\n| Object Class | HOTA | DetA | AssA | LocA |\n|---------------------------|-------|-------|-------|-------|\n| `Person` | 57.96 | 58.99 | 57.11 | 73.29 |\n| `Fourier_GR1_T2_Humanoid` | 60.93 | 60.97 | 60.98 | 74.00 |\n| `Agility_Digit_Humanoid` | 54.00 | 53.32 | 55.32 | 71.02 |\n| `Nova_Carter` | 54.85 | 57.60 | 53.28 | 70.92 |\n| `Transporter` | 33.98 | 38.13 | 32.58 | 67.33 |\n| `Forklift` | 48.05 | 43.78 | 55.20 | 66.45 |\n| Average | 51.63 | 52.13 | 52.41 | 70.50 |\n\n\n## Real-time Inference Performance \n\nInference runs through the DeepStream pipeline on TensorRT with mixed precision (FP16+FP32). The TensorRT columns capture model-only latency, while the DeepStream columns add the instance-bank pre- and post-processing overhead. The table summarizes how many cameras each GPU supports at 30, 15, and 10 FPS for the v2.0 model. Numbers in **bold** are measured (all TensorRT values and certain DeepStream values). DeepStream values not in bold are estimates based on the guidance that the DS microservice adds approximately 30% overhead over Sparse4D TensorRT model performance.\n\n| GPU | TensorRT @30 FPS | DeepStream @30 FPS | TensorRT @15 FPS | DeepStream @15 FPS | TensorRT @10 FPS | DeepStream @10 FPS |\n| :-- | ---------------: | -----------------: | ---------------: | -----------------: | ---------------: | -----------------: |\n| 1 x A100-SXM4-80GB | **13** | 9 | **28** | 19 | **43** | 30 |\n| 1 x B200 | **75** | 52 | **155** | 108 | **200** | 140 |\n| 1 x GB200 | **101** | 70 | **200** | 140 | **200** | 140 |\n| 1 x H100 NVL - 94GB | **23** | 16 | **48** | 33 | **72** | 50 |\n| 1 x H100 SXM HBM3 - 80GB | **28** | 19 | **58** | 40 | **89** | 62 |\n| 1 x H200 | **32** | 22 | **67** | 46 | **103** | 72 |\n| 1 x L4 - 24GB | **3** | 2 | **6** | 4 | **9** | 6 |\n| 1 x L40 | **8** | 5 | **16** | 11 | **26** | 18 |\n| 1 x L40S - 48GB | **9** | 6 | **20** | 14 | **30** | 21 |\n| 1 x RTX 6000 ADA | **9** | 6 | **19** | 13 | **27** | 18 |\n| 1 x RTX PRO 6000 Blackwell (Server) | **28** | 19 | **58** | 40 | **89** | 62 |\n| 1 x RTX PRO 6000 Blackwell (Workstation) | **29** | 20 | **59** | 41 | **90** | 63 |\n| 1 x IGX Thor - T7000 iGPU (no dGPU) | **5** | 3 | **11** | 7 | **17** | 11 |\n| 1 x IGX Thor - T7000 dGPU - RTX PRO 6000 Blackwell Max-Q Workstation Edition | **21** | 14 | **30** | 21 | **30** | 21 |\n| 1 x AGX Thor - T5000 | **4** | 2 | **8** | 5 | **14** | 9 |\n| 1 x DGX Spark | **5** | 3 | **11** | 7 | **17** | 11 |\n\n\n## Changelog\n\n### v2.2\n- Released two backbone variants of the same recipe: ResNet50 (lower latency / smaller footprint) and ResNet101 (high accuracy), shipped under `sparse4d_rn50` and `sparse4d_rn101` NGC model cards respectively.\n- The ResNet50 variant is built on top of an RT-DETR R50 pre-training initialization for stronger image features at half the backbone cost.\n- Added `pallet_truck` to the trained class set (`person`, `gr1_t2`, `agility_digit`, `nova_carter`, `transporter`, `forklift`, `pallet_truck`).\n- Expanded the training corpus to all available MTMC Tracking 2025 scenes (lazy-loaded), 3K+ BEV groups, and refreshed COSMOS-augmented data; mixed FPS training is unchanged.\n- Refreshed kmeans anchor priors per backbone (`_ov_kmeans900_v2.2_r50.npy`, `_ov_kmeans900_v2.2_r101.npy`).\n\n### v2.1\n- Trained on a significantly expanded dataset: more synthetic environments (retail, hospital, supermarket), GTC'24 warehouse data, COSMOS-transferred data with BEV groups, and real-world SCOUT data.\n- Leverages the upgraded camera grouping algorithm (SDU v1.2).\n- Corrected forklift orientation annotations.\n\n### v2.0\n- Increased the number of BEV groups to 1337 to improve accuracy across all benchmarks.\n- Added COSMOS-augmented datasets to the training set, using an 80:20 split between synthetic data and COSMOS augmentations.\n- Introduced multiple FPS settings in training to better handle low-frame-rate streams (1, 2, 3, 5, 6, 10, 15, and 30 FPS).\n- Increased the regression weight for rotation loss to improve bounding box fitting accuracy.\n\n### v1.1\n- Updated input image resolution to 960×540 and introduced camera embeddings to lift maximum camera limits.\n- Expanded training set to 34 scenes spanning 500+ cameras and 380 BEV groups (1–18 cameras each).\n- Supported classes: `Person`, `Fourier_GR1_T2_Humanoid`, `Agility_Digit_Humanoid`, `Nova_Carter`, `Transporter`, `Forklift`.\n- Mixed input frame rates (10, 15, 30 FPS) to improve robustness to lower streaming FPS.\n- Enabled sparser camera grouping with random frame dropping to mirror Data Service frame-loss scenarios.\n- Fixed rotation bug in object annotations and retrained the model with corrected labels.\n- New MSDA FP16 Kernel which helps improve throughput by ~1.5x-2x for Data Center GPUs available in the DeepStream Pipeline inside the Nvidia Video Search and Summarization (VSS) - Warehouse Blueprint release.\n\n### v1.0\n- First commercial warehouse-focused 3D detection and tracking release.\n- Input image resolution: 1408×512 and cameras limited to 20.\n- Supported classes: `Person`, `Fourier_GR1_T2_Humanoid`, `Agility_Digit_Humanoid`, `Nova_Carter`.\n\n\n## How to use this model \n\nIn order to use the model as pre-trained weights for transfer learning, please use the snippet below as a template for the `model` component of the experiment spec file to train a Sparse4D. For more information on experiment spec file, please refer to the [Train Adapt Optimize (TAO) Toolkit User Guide](https://docs.nvidia.com/tao/tao-toolkit/index.html).\n\n```yaml\nmodel:\n type: \"sparse4d\"\n use_grid_mask: true\n use_deformable_func: true\n use_temporal_align: true\n input_shape: [960, 540]\n embed_dims: 256\n neck:\n type: \"FPN\"\n num_outs: 4\n start_level: 0\n out_channels: 256\n in_channels: [256, 512, 1024, 2048]\n add_extra_convs: \"on_output\"\n relu_before_extra_convs: true\n depth_branch:\n type: \"dense_depth\"\n embed_dims: \"${model.embed_dims}\"\n num_depth_layers: 3\n loss_weight: 0.2\n head:\n type: \"sparse4d\"\n num_output: 300\n cls_threshold_to_reg: 0.05\n decouple_attn: true\n return_feature: true\n use_reid_sampling: false\n embed_dims: \"${model.embed_dims}\"\n num_groups: 8\n num_decoder: 6\n num_single_frame_decoder: 1\n drop_out: 0.1\n temporal: true\n with_quality_estimation: true\n instance_bank:\n num_anchor: 900\n anchor: ???\n num_temp_instances: 600\n confidence_decay: 0.8\n feat_grad: false\n default_time_interval: 0.033333\n embed_dims: \"${model.embed_dims}\"\n use_temporal_align: \"${model.use_temporal_align}\"\n anchor_encoder:\n type: 'SparseBox3DEncoder'\n vel_dims: 3\n embed_dims: [128, 32, 32, 64]\n mode: 'cat'\n output_fc: false\n in_loops: 1\n out_loops: 4\n operation_order: [\n \"deformable\", \"ffn\", \"norm\", \"refine\", \"temp_gnn\", \"gnn\", \"norm\", \n \"deformable\", \"ffn\", \"norm\", \"refine\", \"temp_gnn\", \"gnn\", \"norm\", \n \"deformable\", \"ffn\", \"norm\", \"refine\", \"temp_gnn\", \"gnn\", \"norm\", \n \"deformable\", \"ffn\", \"norm\", \"refine\", \"temp_gnn\", \"gnn\", \"norm\", \n \"deformable\", \"ffn\", \"norm\", \"refine\", \"temp_gnn\", \"gnn\", \"norm\", \n \"deformable\", \"ffn\", \"norm\", \"refine\"\n ]\n temp_graph_model:\n type: \"MultiheadAttention\"\n embed_dims: 512\n num_heads: 8\n batch_first: true\n dropout: 0.1\n graph_model:\n type: \"MultiheadAttention\"\n embed_dims: \"${model.head.temp_graph_model.embed_dims}\"\n num_heads: \"${model.head.temp_graph_model.num_heads}\"\n batch_first: true\n dropout: \"${model.head.temp_graph_model.dropout}\"\n norm_layer:\n type: \"LN\"\n normalized_shape: \"${model.embed_dims}\"\n ffn:\n type: \"AsymmetricFFN\"\n in_channels: 512\n pre_norm:\n type: \"LN\"\n embed_dims: 256\n feedforward_channels: 1024\n num_fcs: 2\n ffn_drop: 0.1\n act_cfg:\n type: \"ReLU\"\n inplace: true\n deformable_model:\n embed_dims: \"${model.embed_dims}\"\n num_groups: 8\n num_levels: 4\n attn_drop: 0.15\n use_deformable_func: true\n use_camera_embed: true\n residual_mode: \"cat\"\n kps_generator:\n embed_dims: \"${model.embed_dims}\"\n num_learnable_pts: 6\n fix_scale:\n - [0, 0, 0]\n - [0.45, 0, 0]\n - [-0.45, 0, 0]\n - [0, 0.45, 0]\n - [0, -0.45, 0]\n - [0, 0, 0.45]\n - [0, 0, -0.45]\n refine_layer:\n type: \"SparseBox3DRefinementModule\"\n embed_dims: \"${model.embed_dims}\"\n refine_yaw: true\n with_quality_estimation: true\n sampler:\n num_dn_groups: 5\n num_temp_dn_groups: 3\n dn_noise_scale: [2.0, 2.0, 2.0, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5]\n max_dn_gt: 128\n add_neg_dn: true\n cls_weight: 2.0\n box_weight: 0.25\n reg_weights: [2.0, 2.0, 2.0, 0.5, 0.5, 0.5, 0.0, 0.0, 0.0, 0.0, 0.0]\n use_temporal_align: \"${model.use_temporal_align}\"\n visibility_net:\n type: \"visibility_net\"\n embedding_dim: 256\n hidden_channels: 32\n loss:\n reg:\n type: \"sparse_box_3d\"\n box_weight: 0.25\n cls_allow_reverse: [5, 6, 7]\n cls:\n type: \"focal\"\n use_sigmoid: true\n gamma: 2.0\n alpha: 0.25\n loss_weight: 2.0\n id:\n type: \"cross_entropy_label_smooth\"\n num_ids: \"${dataset.num_ids}\"\n bnneck:\n type: \"bnneck\"\n feat_dim: 256\n num_ids: \"${dataset.num_ids}\"\n decoder:\n type: \"SparseBox3DDecoder\"\n score_threshold: 0.05\n reg_weights: [2.0, 2.0, 2.0, 1 ,1, 1, 1, 1, 1, 1, 1]\n\n```\n\n## Inference & Deployment:\n\nFor deployment of this model, please refer to the our Nvidia Video Search and Summarization (VSS) - Warehouse Blueprint release documentation.\n\n**Acceleration Engine:** Tensor(RT)
\n\n**Test Hardware:**
\n* NVIDIA Datacenter GPUs
\n\n\n### Ethical Considerations \n\nNVIDIA believes Trustworthy AI is a shared responsibility and we have established policies and practices to enable development for a wide array of AI applications. When downloaded or used in accordance with our terms of service, developers should work with their internal model team to ensure this model meets requirements for the relevant industry and use case and addresses unforeseen product misuse. \n\nPlease make sure you have proper rights and permissions for all input image and video content; if image or video includes people, personal health information, or intellectual property, the image or video generated will not blur or maintain proportions of image subjects included. \n\nFor more detailed information on ethical considerations for this model, please see the Model Card++ Explainability, Bias, Safety & Security, and Privacy Sub-cards.\n\nPlease report model quality, risk, security vulnerabilities or NVIDIA AI Concerns [here](nvidia.com/en-us/support/submit-security-vulnerability/).","displayName","Sparse4D (ResNet-101)",["D",1780092955432],"publisher","NVIDIA","latestVersion","versions","collectionCounts","imageUrl","/api/image/nvidia/tao/models/sparse4d_rn101","rawImageUrl","https://dz112fgwz7ogh.cloudfront.net/logos/Nvidia-Centric-TAO.png","canGuestDownload","latestVersionSizeInBytes",804552288,"licenseTerms",{"_324":325,"_326":327,"_328":329,"_330":31},"isMultinodeEnabled","isMultiArchitecture","bias","| Field | Response |\n| -- | -- |\n|Participation considerations from adversely impacted groups [(protected classes)](https://www.senate.ca.gov/content/protected-classes) in model design and testing: | None. |\n| Measures taken to mitigate against unwanted bias: | None. |","explainability","| Field | Response |\n| -- | -- |\n| Intended Application(s) & Domain(s): | 3D Multi-Camera Detection & Tracking of Objects. |\n| Model Type: | Resnet101 Backbone + Transformer |\n| Intended Users: | This model is intended for developers working in industrial applications. |\n| Output: | Bounding box information & object IDs in 3D coordinates |\n| Describe how the model works: | Input images from multiple cameras are encoded & passed onto decoder layers for regression & classification tasks. |\n| Name the adversely impacted groups this has been tested to deliver comparable outcomes: | Not Applicable. |\n| Technical Limitations & Mitigations: | This model may not perform well for objects with poor visibility across all cameras. Model may be fine-tuned to mitigate for accuracy. Additional data can be collected using NVIDIA Omniverse to mitigate poor object visibility across cameras. |\n| Verified to have met prescribed NVIDIA standards: | Yes |\n| Performance Metrics: | Average Precision (AP) & Mean Average Precision (mAP) |\n| Potential Known Risks: | This model may produce incorrectly placed bounding boxes and tracking IDs, particularly in instances with poor visibility. This model was trained primarily on synthetic data from warehouse scenes. As a result, it may produce false positives and false negatives when applied to different scenes, domains, or object categories outside its training scope. | \n| Licensing: | GOVERNING TERMS: Use of this model is governed by the [NVIDIA Open Model License](https://www.nvidia.com/en-us/agreements/enterprise-software/nvidia-open-model-license/). |","privacy","Field | Response\n:----------------------------------------------------------------------------------------------------------------------------------|:-----------------------------------------------\nGeneratable or reverse engineerable personal data? | No\nPersonal data used to create this model? | No\nHow often is dataset reviewed? | Before Release\nIs there provenance for all datasets used in training? | Yes\nDoes data labeling (annotation, metadata) comply with privacy laws? | Yes\nIs data compliant with data subject requests for data correction or removal, if such a request was made? | No, not possible with externally-sourced data.\nApplicable Privacy Policy | https://www.nvidia.com/en-us/about-nvidia/privacy-policy/ \nWas data from user interactions with the AI model (e.g. user input and prompts) used to train the model? | No","safetyAndSecurity","| Field | Response |\n| -- | -- |\n| Model Application(s): | 3D Multi-Camera Detection & Tracking of Objects |\n| Describe the life-critical application (if present). | This model performs 3D Detection & Tracking for multiple objects, including persons, humanoids, and autonomous mobile robots. This model may be used in the NVIDIA VSS release, for the purposes of providing real time alerts for events such as potential safety violations and security breaches, identifying on-demand insights pertaining to space and asset utilization, and ensuring the continuous integration of data for process optimization and strategic planning. |\n| Use Case Restrictions: | GOVERNING TERMS: Use of this model is governed by the [NVIDIA Open Model License](https://www.nvidia.com/en-us/agreements/enterprise-software/nvidia-open-model-license/). |\n| Describe access restrictions (if any): | The Principle of least privilege (PoLP) is applied limiting access for dataset generation and model development. Restrictions enforce dataset access during training, and dataset license constraints adhered to. |","featureLabels",[],"displayLabels",[286,287,288,289,290,291,292,293,294,295],"policyLabels",[],"systemLabels",[],"products",[],"hasNimMetadata",{"_248":320,"_297":321,"_298":322,"_300":301},{"_248":318,"_297":318,"_298":319,"_300":301},{"_248":316,"_297":316,"_298":317,"_300":301},{"_248":314,"_297":314,"_298":315,"_300":301},{"_248":312,"_297":312,"_298":313,"_300":301},{"_248":310,"_297":310,"_298":311,"_300":301},{"_248":308,"_297":308,"_298":309,"_300":301},{"_248":306,"_297":306,"_298":307,"_300":301},{"_248":304,"_297":304,"_298":305,"_300":301},{"_248":296,"_297":296,"_298":299,"_300":301},"TAO Toolkit","value","filter",{"_302":303,"_297":296},"category","label","field","framework","TAO",{"_302":301,"_297":304},"Object Detection and Tracking in 3D",{"_302":301,"_297":306},"NSPECT-XT56-057L",{"_302":301,"_297":308},"NSPECT-FY5H-V6CO",{"_302":301,"_297":310},"Multi-Camera Network",{"_302":301,"_297":312},"False",{"_302":301,"_297":314},"Deep Learning",{"_302":301,"_297":316},"CV",{"_302":301,"_297":318},"Computer Vision","soln_computer_vision",{"_302":323,"_297":321},"technology","governingTerms","

Governing Terms: Use of this model is governed by the NVIDIA Open Model License.

","licenseId","NVIDIA proprietary","licenseVersion","v1.0","needsAcceptance",{"_119":1306,"_121":1307},"ngcApps",[1259],"csaApiBaseUrl","https://www.nvidia.com/ai-agents/csa-agent/dcsm","datadogRum",{"_1240":1241,"_1242":1243,"_1244":1245,"_1246":1247,"_1248":342,"_151":368,"_1249":1250,"_1251":55,"_1252":98,"_1253":98,"_1254":1255,"_1256":98,"_1257":1258},"amplitude",{"_1237":1238,"_880":-7,"_1239":31},"disableJs","environment","prod","featureFlagsBlob","JTdCJTIydGVzdEZsYWclMjIlM0FmYWxzZSUyQyUyMnNob3dTZWFyY2hMYWJlbFNldEZpbHRlcnMlMjIlM0FmYWxzZSUyQyUyMmVuYWJsZUNzYVdpZGdldCUyMiUzQWZhbHNlJTJDJTIyZW5hYmxlUHJvZHVjdEdyb3VwaW5nQnlTdWJzY3JpcHRpb24lMjIlM0FmYWxzZSUyQyUyMmNvbnRhaW5lclNjYW5uaW5nUmVkZXNpZ25NMiUyMiUzQWZhbHNlJTdE","featureFlagsGuarded","isDevLikeEnv","isPlaywright","isMobileDevice","categorizedOrgs",{"_1231":1232,"_1233":1234,"_1235":1236},"currentOrgContext","currentContextRoles",[],"enterpriseOrgs",[],"manageableExternalUserGroups",[],"ngcApiOrigin","ngcHelmOrigin","origin","https://catalog.ngc.nvidia.com","usesMocks","theme","year",2026,"now",1787683775129,"1.12.0","defaultSearchSuggestions",{"_1120":181,"_1121":1122,"_1123":1124,"_1125":1126},"announcements",[],"consentState",{"_406":1110,"_1111":98,"_1112":31,"_1113":31,"_1114":1115},"consentConfig",{"_388":389,"_390":391,"_392":393,"_394":395,"_396":397,"_398":399,"_400":401,"_402":403,"_404":405,"_406":407},"showContactAdmin",1787683775135,[],[],[382,383,384,385,386,387],{"_45":59,"_47":70,"_49":114,"_51":-5,"_53":-5,"_54":55},{"_45":59,"_47":68,"_49":114,"_51":-5,"_53":-5,"_54":55},{"_45":59,"_47":66,"_49":114,"_51":-5,"_53":-5,"_54":55},{"_45":59,"_47":60,"_49":114,"_51":-5,"_53":-5,"_54":55},{"_45":56,"_47":57,"_49":114,"_51":-5,"_53":-5,"_54":55},{"_45":46,"_47":48,"_49":114,"_51":-5,"_53":-5,"_54":55},"mainInfoText","We and our third-party partners (including social media, advertising, and analytics partners) use cookies and other tracking technologies to collect, store, monitor, and process certain information about you when you visit our website. The information collected might relate to you, your preferences, or your device. We use that information to make the site work, analyze performance and traffic on our website, provide a more personalized web experience, and assist in our marketing efforts.

Under certain privacy laws, you have the right to direct us not to \"sell\" or \"share\" your personal information for targeted advertising. To opt-out of the \"sale\" and \"sharing\" of personal information through cookies, you must opt-out of optional cookies using the toggles below. To opt out of the \"sale\" and \"sharing\" of data collected by other means (e.g., online forms) you must also update your data sharing preferences through the NVIDIA Preference Center.

Click on the different category headings below to find out more and change the settings according to your preference. You cannot opt out of Required Cookies as they are deployed to ensure the proper functioning of our website (such as prompting the cookie banner and remembering your settings, etc.). By clicking \"Save and Accept\" or \"Decline All\" at the bottom, you consent to the use of cookies and other tools as described in our Cookie Policy in accordance with your settings and accept our Terms of Service (which contains important waivers). For more information about our privacy practices, please see our Privacy Policy.","mainText","Cookie Settings","alertNoticeText",{"_1102":1103,"_1104":1105,"_1106":1107,"_1108":1109},"bannerAcceptLabel","Accept All","bannerRejectLabel","Reject Optional","bannerSettingsLabel","Manage Settings","preferenceSaveLabel","Save Choices","preferenceRejectLabel","Decline All","alwaysActiveText","Always Active","groups",[408,409,410,411,412],{"_413":792,"_80":793,"_180":794,"_417":98,"_418":18,"_420":795,"_422":796},{"_413":695,"_80":696,"_180":697,"_417":31,"_418":16,"_420":698,"_422":699},{"_413":603,"_80":604,"_180":605,"_417":31,"_418":58,"_420":606,"_422":607},{"_413":424,"_80":425,"_180":426,"_417":31,"_418":427,"_420":428,"_422":429},{"_413":414,"_80":415,"_180":416,"_417":31,"_418":419,"_420":421,"_422":423},"optanonGroupId","C0005","Social Media Cookies","These cookies are set by a range of social media services that we have added to the site to enable you to share our content with your friends and networks. They are capable of tracking your browser across other sites and building up a profile of your interests. This may impact the content and messages you see on other websites you visit. If you do not allow these cookies you may not be able to use or see these sharing tools.","alwaysActive","order",11,"cookies",[],"hosts",[],"C0004","Advertising Cookies","These cookies record your visit to our websites, the pages you have visited and the links you have followed to influence the advertisements that you see on other websites. These cookies and the information they collect may be managed by other companies, including our advertising partners, and may be used to build a profile of your interests and show you relevant advertising on other sites. We and our advertising partners will use this information to make our websites and the advertising displayed on it, more relevant to your interests.",5,[545,546,547,548,549,550,551,552,553,554,555,556,557,558,559,560,561,562,563,564,565,566],[430,431,432,433,434,435,436,437,438,439,440,441,442,443,444,445,446],{"_447":540,"_248":540,"_180":181,"_420":541},{"_447":536,"_248":536,"_180":181,"_420":537},{"_447":531,"_248":531,"_180":181,"_420":532},{"_447":527,"_248":527,"_180":181,"_420":528},{"_447":522,"_248":522,"_180":181,"_420":523},{"_447":517,"_248":517,"_180":181,"_420":518},{"_447":509,"_248":509,"_180":181,"_420":510},{"_447":501,"_248":501,"_180":181,"_420":502},{"_447":496,"_248":496,"_180":181,"_420":497},{"_447":491,"_248":491,"_180":181,"_420":492},{"_447":486,"_248":486,"_180":181,"_420":487},{"_447":481,"_248":481,"_180":181,"_420":482},{"_447":476,"_248":476,"_180":181,"_420":477},{"_447":469,"_248":469,"_180":181,"_420":470},{"_447":464,"_248":464,"_180":181,"_420":465},{"_447":457,"_248":457,"_180":181,"_420":458},{"_447":448,"_248":448,"_180":181,"_420":449},"hostName","amazon-adsystem.com",[450,451],{"_80":456,"_453":448,"_454":31,"_180":455},{"_80":452,"_453":448,"_454":31,"_180":455},"ad-privacy","host","isSession","This domain is owned by online retailer Amazon and is used as part of its affiliate marketing programme.","ad-id","doubleclick.net",[459,460],{"_80":463,"_453":457,"_454":31,"_180":462},{"_80":461,"_453":457,"_454":31,"_180":462},"IDE","This domain is owned by Doubleclick (Google). The main business activity is: Doubleclick is Googles real time bidding advertising exchange","test_cookie","bing.com",[466],{"_80":467,"_453":464,"_454":31,"_180":468},"MUID","This domain is owned by Mircosoft - it is the site for the search engine Bing.","assets.adobedtm.com",[471,472],{"_80":475,"_453":469,"_454":31,"_180":474},{"_80":473,"_453":469,"_454":31,"_180":474},"kndctr_F207D74D549850760A4C98C6_AdobeOrg_*","If you know anything about this host and its uses please add your contributions here. ","kndctr_F207D74D549850760A4C98C6_AdobeOrg_consent","tiktok.com",[478],{"_80":479,"_453":476,"_454":31,"_180":480},"_ttp","TikTok is a social media platform owned by ByteDance, known for short-form video sharing. The service is used to create and share videos, and it utilizes cookies for various purposes including advertising and analytics.","t.co",[483],{"_80":484,"_453":481,"_454":31,"_180":485},"muc_ads","This domain is owned by Twitter. It is used as a URL shortening service, which allows Twitter to collect data on how links are shared.","snowplow.com",[488],{"_80":489,"_453":486,"_454":31,"_180":490},"sp","Testing","google.com",[493],{"_80":494,"_453":491,"_454":31,"_180":495},"NID","This domain is owned by Google Inc. Although Google is primarily known as a search engine, the company provides a diverse range of products and services. Its main source of revenue however is advertising. Google tracks users extensively both through its own products and sites, and the numerous technologies embedded into many millions of websites around the world. It uses the data gathered from most of these services to profile the interests of web users and sell advertising space to organisations based on such interest profiles as well as aligning adverts to the content on the pages where its customer's adverts appear.","linkedin.com",[498],{"_80":499,"_453":496,"_454":31,"_180":500},"bcookie","This domain is owned by LinkedIn, the business networking platform. It typically acts as a third party host where website owners have placed one of its content sharing buttons in their pages, although its content and services can be embedded in other ways. Although such buttons add functionality to the website they are on, cookies are set regardless of whether or not the visitor has an active Linkedin profile, or agreed to their terms and conditions. For this reason it is classified as a primarily tracking/targeting domain.","twitter.com",[503,504],{"_80":507,"_453":501,"_454":31,"_180":508},{"_80":505,"_453":501,"_454":31,"_180":506},"personalization_id","This domain is owned by Twitter. The main business activity is: Social Networking Services. Where twitter acts as a third party host, it collects data through a range of plug-ins and integrations, that is primarily used for tracking and targeting.","guest_*","This domain is owned by X (formerly Twitter). The main business activity is: Social Networking Services. Where X (formerly Twitter) acts as a third party host, it collects data through a range of plug-ins and integrations, that is primarily used for tracking and targeting.","youtube.com",[511,512],{"_80":515,"_453":509,"_454":31,"_180":516},{"_80":513,"_453":509,"_454":31,"_180":514},"VISITOR_INFO1_LIVE","This cookie is used as a unique identifier to track viewing of videos","__Secure-*","YouTube is a Google owned platform for hosting and sharing videos. YouTube collects user data through videos embedded in websites, which is aggregated with profile data from other Google services in order to display targeted advertising to web visitors across a broad range of their own and other websites.","demdex.net",[519],{"_80":520,"_453":517,"_454":31,"_180":521},"demdex","This cookie helps Adobe Audience Manger perform basic functions such as visitor identification, ID synchronization, segmentation, modeling, reporting, etc.","im-apps.net",[524],{"_80":525,"_453":522,"_454":31,"_180":526},"receive-cookie-deprecation","This domain is owned by im-apps.net which provides information and resources about various instant messaging (IM) applications to help users make informed decisions based on their communication needs.","influ2.com",[529],{"_80":530,"_453":527,"_454":31,"_180":490},"R","6sc.co",[533],{"_80":534,"_453":531,"_454":31,"_180":535},"6suuid","This domain is associated with the company 6Sense Insights, Inc., which provides predictive analytics and account engagement services. It is primarily used for tracking and targeting in marketing automation and advertising.","hushly.com",[538],{"_80":539,"_453":536,"_454":31,"_180":490},"_hly_gvid","t.contentsquare.net",[542],{"_80":543,"_453":540,"_454":98,"_180":544},"_cs_*","This is a session cookie. This is a pattern type cookie where a random 13-digit number is appended to the prefix _cs_.",{"_80":601,"_453":568,"_454":31,"_180":602},{"_80":599,"_453":568,"_454":31,"_180":600},{"_80":597,"_453":568,"_454":31,"_180":598},{"_80":479,"_453":568,"_454":31,"_180":596},{"_80":573,"_453":568,"_454":98,"_180":575},{"_80":595,"_453":568,"_454":31,"_180":181},{"_80":594,"_453":571,"_454":31,"_180":181},{"_80":592,"_453":568,"_454":31,"_180":593},{"_80":590,"_453":568,"_454":31,"_180":591},{"_80":589,"_453":568,"_454":31,"_180":181},{"_80":587,"_453":568,"_454":31,"_180":588},{"_80":573,"_453":586,"_454":98,"_180":575},{"_80":584,"_453":571,"_454":31,"_180":585},{"_80":584,"_453":574,"_454":31,"_180":585},{"_80":543,"_453":568,"_454":31,"_180":544},{"_80":582,"_453":568,"_454":31,"_180":583},{"_80":489,"_453":580,"_454":31,"_180":581},{"_80":578,"_453":568,"_454":31,"_180":579},{"_80":576,"_453":568,"_454":31,"_180":577},{"_80":573,"_453":574,"_454":98,"_180":575},{"_80":570,"_453":571,"_454":31,"_180":572},{"_80":567,"_453":568,"_454":98,"_180":569},"AMCVS_xxxxxAdobeOrg","nvidia.com","This is a pattern type cookie name associated with Adobe Marketing Cloud. It stores a unique visitor identifier, and uses an organisation identifier to allow a company to track users across their domains and services.This is a pattern type cookie name associated with Adobe Marketing Cloud. It stores a unique visitor identifier, and uses an organisation identifier.","_rdt_pn","www.nvidia.com","This cookie name is associated with www.mind.org.uk which is used for marketing purposes.","AMP_*","build.nvidia.com","Supports Amplitude product analytics.","_gcl_au","Used by Google AdSense for experimenting with advertisement efficiency across websites using their services","_mkto_trk","This cookie is associated with an email marketing service provided by Marketo. This tracking cookie allows a website to link visitor behaviour to the recipient of an email marketing campaign, to measure campaign effectiveness.","investor.nvidia.com","This cookie is associated with Eventbrite and is used to deliver content tailored to the end user's interests and improve content creation. This cookie is also used for event-booking purposes.","_gat_gtag_UA_140825383_2","Google Analytics","_rdt_uuid","This cookie is set by Reddit and is used for tracking the views of embedded videos and performance of the advertisements.","catalog.ngc.nvidia.com","_uetsid","This cookie is used by Bing to determine what ads should be shown that may be relevant to the end user perusing the site. ","kndctr_F207D74D549850760A4C98C6_AdobeOrg_bc_session_id","ttcsid_XXXXXXXXXXXXXXXXXXXXCategory","This cookie name is associated with TikTok which is used to track your activity on the platform, including how you interact with their embedded services on other websites.This cookie is associated with TikTok. It helps track your activity on the platform, including how you interact with their embedded services on other websites.","_fbp","Used by Facebook to deliver a series of advertisement products such as real time bidding from third party advertisers","_zitok","_twpid","This cookie is typically associated with the use of Taptap Digital, a company specializing in digital advertising. It is used for tracking and ad personalization purposes.","_tt_enable_cookie","This cookie name is associated with TikTok which is used to collect data about behavior and purchases on the website and to measure the effect of advertising. It is also used to track website activity to help optimize advertising efforts.","AMCV_xxxxxAdobeOrg","This cookie is used to identify a unique visitor for tracking purposes by Adobe Experience Cloud.This is a pattern type cookie name associated with Adobe Marketing Cloud. It stores a unique visitor identifier, and uses an organisation identifier to allow a company to track users across their domains and services.","ttcsid","This cookie name is associated with TikTok which is used to enable TikTok Ads session tracking by assigning a unique session ID to users who click on TikTok ads, allowing for the attribution of user behavior and conversions within the same session.This cookie is associated with TikTok. It helps track your activity on the platform, including how you interact with their embedded services on other websites.","C0003","Personalization Cookies","These cookies collect data about how you have interacted with our website to help us improve your web experience, such as which pages you have visited. These cookies may store a unique ID so that our system will remember you when you return. They may be set by us or by third party providers whose services we have added to our pages. These cookies enable us to provide enhanced website functionality and personalization as well as make the marketing messages we send to you more relevant to your interests. If you do not allow these cookies, then some or all of these services may not function properly.",[637,638,639,640,641,642,643,644,645,646,647,648,649,650,651,652,653,654,655,656,657,658,659,660,661],[608,609,610,611,612],{"_447":522,"_248":522,"_180":181,"_420":632},{"_447":496,"_248":496,"_180":181,"_420":627},{"_447":622,"_248":622,"_180":181,"_420":623},{"_447":617,"_248":617,"_180":181,"_420":618},{"_447":613,"_248":613,"_180":181,"_420":614},"nvidia.partners",[615],{"_80":616,"_453":613,"_454":31,"_180":490},"edge","app.hushly.com",[619],{"_80":620,"_453":617,"_454":31,"_180":621},"*CORS","This domain is associated with Hushly, a platform that provides lead conversion solutions for content marketing by using personalization and content engagement strategies.","github.com",[624],{"_80":625,"_453":622,"_454":31,"_180":626},"_octo","This domain is owned by Github. The main business activity is: Source code repository and collaboration platform",[628,629],{"_80":631,"_453":496,"_454":31,"_180":500},{"_80":630,"_453":496,"_454":31,"_180":500},"lidc","li_gc",[633,634],{"_80":636,"_453":522,"_454":31,"_180":526},{"_80":635,"_453":522,"_454":31,"_180":526},"imid_*","imid",{"_80":662,"_453":571,"_454":31,"_180":664},{"_80":693,"_453":568,"_454":31,"_180":694},{"_80":662,"_453":692,"_454":31,"_180":664},{"_80":690,"_453":568,"_454":98,"_180":691},{"_80":688,"_453":568,"_454":31,"_180":689},{"_80":687,"_453":670,"_454":98,"_180":181},{"_80":662,"_453":666,"_454":31,"_180":664},{"_80":685,"_453":568,"_454":31,"_180":686},{"_80":662,"_453":574,"_454":31,"_180":664},{"_80":673,"_453":571,"_454":31,"_180":181},{"_80":662,"_453":684,"_454":31,"_180":664},{"_80":682,"_453":568,"_454":31,"_180":683},{"_80":679,"_453":680,"_454":31,"_180":681},{"_80":677,"_453":678,"_454":31,"_180":181},{"_80":662,"_453":674,"_454":31,"_180":664},{"_80":676,"_453":568,"_454":31,"_180":181},{"_80":662,"_453":675,"_454":31,"_180":664},{"_80":673,"_453":674,"_454":31,"_180":181},{"_80":672,"_453":670,"_454":98,"_180":181},{"_80":662,"_453":671,"_454":31,"_180":664},{"_80":662,"_453":670,"_454":31,"_180":664},{"_80":662,"_453":669,"_454":31,"_180":664},{"_80":668,"_453":568,"_454":98,"_180":664},{"_80":665,"_453":666,"_454":98,"_180":667},{"_80":662,"_453":663,"_454":31,"_180":664},"_gd_*","info.nvidia.com","This cookie contains no identifying information and is used for diagnostic purposes by Impact Radius, a service we use to track signup commissions for our affiliate program. We have categorized it under Functionality because it is necessary for the operation of our affiliate program.","nv_country_code","developer.nvidia.com","This cookie is likely used to store the country code of the visitor to provide localized content and possibly for legal compliance with regional data privacy laws.","_gdxxxxxxxxxxxxx","nvidianews.nvidia.com","docs.api.nvidia.com","research.nvidia.com","readme_library","wcmmode","ai-podcast.nvidia.com","resources.nvidia.com","geo_locator","openedx-language-preference","learn.nvidia.com","doxygen_width","raytracing-docs.nvidia.com","This cookie is likely used to remember the width preferences of a sidebar or interface element within a documentation platform like Doxygen.","_sfid_eb78","This cookie is categorized as Functionality Cookie.","crowdsource.nvidia.com","mbox","Cookie set by Adobe SiteCatalyst analytics software - used for measuring the performance of page content using A/B split testing","readme_language","vid","This cookie name 'vid' is used for tracking the search request made by the user on the 'Find a practitioner' page.","at_check","Used by Adobe Target to check if cookies are enabled/supported on the browser","blogs.nvidia.com","nvHeadless","This cookie is likely used to track if a user is using the site in a 'headless' non-GUI mode, which could be for accessibility reasons or for automated processes.","C0002","Performance Cookies","These cookies are used to provide quantitative measures of our website visitors, such as the number of times you visit, time on page, your mouse movements, scrolling, clicks and keystroke activity on the websites; other browsing, search, or product research behavior; and what brought you to our site. These cookies may store a unique ID so that our system will remember you when you return. Information collected with these cookies is used to measure and find ways to improve website performance.",[720,721,722,723,724,725,726,727,728,729,730,731,732,733,734,735,736,737,738,739,740,741,742,743,744,745,746,747,748],[700,701,702,703],{"_447":715,"_248":715,"_180":181,"_420":716},{"_447":710,"_248":710,"_180":181,"_420":711},{"_447":509,"_248":509,"_180":181,"_420":707},{"_447":622,"_248":622,"_180":181,"_420":704},[705],{"_80":706,"_453":622,"_454":98,"_180":626},"_gh_sess",[708],{"_80":709,"_453":509,"_454":98,"_180":516},"YSC","bizible.com",[712],{"_80":713,"_453":710,"_454":31,"_180":714},"_BUID","Bizible is a marketing attribution software that tracks the performance of marketing efforts and connects them to revenue, primarily used for B2B environments.","a.usbrowserspeed.com",[717],{"_80":718,"_453":715,"_454":31,"_180":719},"tuid","This domain is owned by Ookla which provides internet speed testing services through its Speedtest application.",{"_80":790,"_453":568,"_454":31,"_180":791},{"_80":775,"_453":789,"_454":98,"_180":776},{"_80":758,"_453":571,"_454":31,"_180":683},{"_80":758,"_453":671,"_454":31,"_180":683},{"_80":787,"_453":568,"_454":31,"_180":788},{"_80":775,"_453":574,"_454":98,"_180":776},{"_80":785,"_453":568,"_454":31,"_180":786},{"_80":775,"_453":571,"_454":98,"_180":776},{"_80":783,"_453":670,"_454":31,"_180":784},{"_80":775,"_453":692,"_454":98,"_180":776},{"_80":755,"_453":571,"_454":31,"_180":757},{"_80":781,"_453":568,"_454":98,"_180":782},{"_80":779,"_453":568,"_454":31,"_180":780},{"_80":777,"_453":568,"_454":98,"_180":778},{"_80":775,"_453":666,"_454":98,"_180":776},{"_80":773,"_453":568,"_454":31,"_180":774},{"_80":771,"_453":568,"_454":31,"_180":772},{"_80":769,"_453":568,"_454":98,"_180":770},{"_80":767,"_453":675,"_454":31,"_180":768},{"_80":755,"_453":666,"_454":31,"_180":757},{"_80":765,"_453":568,"_454":31,"_180":766},{"_80":763,"_453":568,"_454":98,"_180":764},{"_80":761,"_453":568,"_454":31,"_180":762},{"_80":759,"_453":568,"_454":31,"_180":760},{"_80":758,"_453":568,"_454":31,"_180":683},{"_80":755,"_453":756,"_454":31,"_180":757},{"_80":753,"_453":568,"_454":31,"_180":754},{"_80":751,"_453":568,"_454":98,"_180":752},{"_80":749,"_453":568,"_454":31,"_180":750},"_ga_xxxxxxxxxx","Used by Google Analytics to identify and track an individual session with your device.","s_ppvl","This cookie is set by Adobe Analytics to store information on the percentage of the page displayed to the end user.","coveo_visitorId","This cookie name is associated with Coveo which is used to store a unique user ID for marketing purposes and to track the number of distinct users who submitted a query or clicked an item, counting them only once even if they did the same query or clicked the same item over the course of many visits.","_dd_s","marketplace.nvidia.com","Datadog Cookie used to group all events generated from a unique user session across multiple pages. It contains the current session ID, whether the session is excluded due to sampling, and the expiration date of the session. The cookie is extended for an extra 15 minutes every time the user interacts with the website, up to the maximum user session duration (4 hours).","nvweb_A","_ga","This cookie name is associated with Google Universal Analytics - which is a significant update to Google's more commonly used analytics service. This cookie is used to distinguish unique users by assigning a randomly generated number as a client identifier. It is included in each page request in a site and used to calculate visitor, session and campaign data for the sites analytics reports. By default it is set to expire after 2 years, although this is customisable by website owners.","_uetvid","This is a cookie utilised by Microsoft Bing Ads and is a tracking cookie. It allows us to engage with a user that has previously visited our website.","tp","This cookie is associated with Adobe Analytics site tracking. It registers if the website's scroll-depth detection is active. This function remembers how far the user has scrolled on the current session across the website's sub-pages.","_hly_vid","This cookie _hly_vid is used to record website and server performance with each visit to the site. The data is aggregated and analyzed; therefore, it remains anonymous upon use.","_session_id","This cookie creates an interim session ID used as an in-session user ID.\n","s_cc","This cookie is associated with the Adobe Site Catalyst. It determines whether cookies are enabled in the web browser.","s_ecid","Adobe Analytics cookie containing the the Adobe Experience Cloud 1st party visitor id if the Analytics AMCV visitor id is not found.","_hjSessionUser_xxxxxx","Hotjar cookie that is set when a user first lands on a page with the Hotjar script. It is used to persist the Hotjar User ID, unique to that site on the browser. This ensures that behavior in subsequent visits to the same site will be attributed to the same user ID.","_hjHasCachedUserAttributes","This cookie name is associated with Hotjar which is used to know whether the data set in _hjUserAttributes Local Storage item is up to date or not.","s_ppv","This cookie is associated with Adobe Analytics site tracking and is used to measure a user's scroll activity to see how much of a page they view before moving to another page.","_gid","This cookie name is associated with Google Universal Analytics. This appears to be a new cookie and as of Spring 2017 no information is available from Google. It appears to store and update a unique value for each page visited.","_hjTLDTest","When the Hotjar script executes we try to determine the most generic cookie path we should use, instead of the page hostname. This is done so that cookies can be shared across subdomains (where applicable). To determine this, we try to store the _hjTLDTest cookie for different URL substring alternatives until it fails. After this check, the cookie is removed.","ekfls-2025-03-27","This cookie name is associated with readme.io which is used to help maintain data for the API reference support page.","s_getNewRepeat","This cookie is used to determine whether a visitor is a new visitor or a repeat visitor to the site, and is used for analytical purposes.","_hly_sid","This cookie is likely associated with website analytics, as the prefix '_hly_' suggests a connection to Holistics, a business intelligence and analytics platform.","docs.nvidia.com","_hjSession_xxxxxx","A cookie that holds the current session data. This ensues that subsequent requests within the session window will be attributed to the same Hotjar session.","C0001","Required Cookies","These cookies enable core functionality such as security, network management, and accessibility. These cookies are required for the site to function and cannot be turned off.",[937,938,939,940,941,942,943,944,945,946,947,948,949,950,951,952,953,954,955,956,957,958,959,960,961,962,963,964,965,966,967,968,969,970,971,972,973,974,975,976,977,978,979,980,981,982,983,984,985,986,987,988,989,990,991,992,993,994,995,996,997,998,999,1000,1001,1002,1003,1004,1005,1006,1007,1008,1009],[797,798,799,800,801,802,803,804,805,806,807,808,809,810,811,812,813,814,815,816,817,818,819,820,821],{"_447":934,"_248":934,"_180":181,"_420":935},{"_447":928,"_248":928,"_180":181,"_420":929},{"_447":925,"_248":925,"_180":181,"_420":926},{"_447":913,"_248":913,"_180":181,"_420":914},{"_447":910,"_248":910,"_180":181,"_420":911},{"_447":907,"_248":907,"_180":181,"_420":908},{"_447":902,"_248":902,"_180":181,"_420":903},{"_447":509,"_248":509,"_180":181,"_420":899},{"_447":896,"_248":896,"_180":181,"_420":897},{"_447":501,"_248":501,"_180":181,"_420":894},{"_447":496,"_248":496,"_180":181,"_420":892},{"_447":887,"_248":887,"_180":181,"_420":888},{"_447":882,"_248":882,"_180":181,"_420":883},{"_447":877,"_248":877,"_180":181,"_420":878},{"_447":874,"_248":874,"_180":181,"_420":875},{"_447":622,"_248":622,"_180":181,"_420":871},{"_447":481,"_248":481,"_180":181,"_420":868},{"_447":617,"_248":617,"_180":181,"_420":861},{"_447":613,"_248":613,"_180":181,"_420":855},{"_447":849,"_248":849,"_180":181,"_420":850},{"_447":844,"_248":844,"_180":181,"_420":845},{"_447":839,"_248":839,"_180":181,"_420":840},{"_447":835,"_248":835,"_180":181,"_420":836},{"_447":830,"_248":830,"_180":181,"_420":831},{"_447":822,"_248":822,"_180":181,"_420":823},"polyfill.io",[824,825],{"_80":828,"_453":822,"_454":31,"_180":829},{"_80":826,"_453":822,"_454":31,"_180":827},"cf_ob_info","This cookie is associated with sites using CloudFlare. It is used to improve page load times and to override any security restrictions based on the visitors IP address. It does not contain any user identification information.","cf_use_ob","This cookie is associated with sites using CloudFlare. It is used to improve page load times and to override any security restrictions based on the visitors IP address. It does not contain any user identification information. ","www.recaptcha.net",[832],{"_80":833,"_453":830,"_454":31,"_180":834},"_GRECAPTCHA","This domain is used by Google reCAPTCHA, which provides spam protection services to websites by detecting if a user is a human or a bot.","api2.hcaptcha.com",[837],{"_80":838,"_453":835,"_454":31,"_180":490},"__cflb","instagram.com",[841],{"_80":842,"_453":839,"_454":31,"_180":843},"csrftoken","Instagram is a social networking service owned by Facebook, Inc. It's designed for sharing photos and videos from a smartphone. The service also uses cookies for various purposes, including analytics, personalization, and advertising.","readme.io",[846],{"_80":847,"_453":844,"_454":31,"_180":848},"__cf_bm","The __cf_bm cookie is a cookie necessary to support Cloudflare Bot Management, currently in private beta. As part of our bot management service, this cookie helps manage incoming traffic that matches criteria associated with bots.","nvidiacorp.us-5.evergage.com",[851,852],{"_80":854,"_453":849,"_454":31,"_180":490},{"_80":853,"_453":849,"_454":31,"_180":490},"AWSALBTG","AWSALBTGCORS",[856,857],{"_80":859,"_453":613,"_454":31,"_180":860},{"_80":858,"_453":613,"_454":31,"_180":490},"ak_bmsc","bm_mi","This domain is likely associated with NVIDIA's partner program, which could involve marketing, sales, and support services for partners.",[862,863,864],{"_80":867,"_453":617,"_454":31,"_180":490},{"_80":866,"_453":617,"_454":98,"_180":490},{"_80":865,"_453":617,"_454":31,"_180":490},"AWSALBCORS","JSESSIONID","AWSALB",[869],{"_80":847,"_453":481,"_454":31,"_180":870},"This is a CloudFoundry cookie",[872],{"_80":873,"_453":622,"_454":31,"_180":626},"logged_in","hcaptcha.com",[876],{"_80":847,"_453":874,"_454":31,"_180":870},"tracking.intentsify.io",[879],{"_80":880,"_453":877,"_454":31,"_180":881},"userId","Used to track usage of the embedded video player.","static.reo.dev",[884],{"_80":885,"_453":882,"_454":31,"_180":886},"cookietest","Common cookie name could have a number of different origins. Where this is first party and a session cookie, its most likely to do with checking to see if the browser is set to block or allow cookies.","www.datadoghq-browser-agent.com",[889],{"_80":890,"_453":887,"_454":31,"_180":891},"dd_cookie_test_","dd_cookie_test",[893],{"_80":847,"_453":496,"_454":31,"_180":870},[895],{"_80":847,"_453":501,"_454":31,"_180":870},"supabase.co",[898],{"_80":847,"_453":896,"_454":31,"_180":870},[900],{"_80":901,"_453":509,"_454":31,"_180":516},"VISITOR_PRIVACY_METADATA","dosa-prd.azurewebsites.net",[904],{"_80":905,"_453":902,"_454":31,"_180":906},"access_token","This cookie is used to obtain the permissions for users of the portal to allow for API access in order to retrieve information required to perform portal functions.","w.hcaptcha.com",[909],{"_80":847,"_453":907,"_454":31,"_180":870},"cdn.readme.io",[912],{"_80":847,"_453":910,"_454":31,"_180":870},"queue-it.net",[915,916,917,918,919],{"_80":924,"_453":913,"_454":31,"_180":181},{"_80":923,"_453":913,"_454":31,"_180":181},{"_80":922,"_453":913,"_454":31,"_180":181},{"_80":921,"_453":913,"_454":31,"_180":181},{"_80":920,"_453":913,"_454":31,"_180":181},"Queue-it-{queueId}","Queue-it-{dmonvidia }________{eventId}","Queue-it-token","queueitsoftblock_{queueId}","Queue-it-{nvidia}________{eventId}","x.com",[927],{"_80":847,"_453":925,"_454":31,"_180":848},"api.hcaptcha.com",[930,931],{"_80":933,"_453":928,"_454":31,"_180":490},{"_80":838,"_453":928,"_454":31,"_180":932},"This domain is associated with hCaptcha, a service that provides CAPTCHA and reCAPTCHA alternatives for user verification and bot protection.","hmt_id","vimeo.com",[936],{"_80":847,"_453":934,"_454":31,"_180":870},{"_80":1070,"_453":568,"_454":31,"_180":1101},{"_80":1099,"_453":663,"_454":98,"_180":1100},{"_80":1092,"_453":1061,"_454":98,"_180":1093},{"_80":1098,"_453":756,"_454":98,"_180":181},{"_80":1096,"_453":568,"_454":31,"_180":1097},{"_80":854,"_453":1082,"_454":31,"_180":1021},{"_80":1029,"_453":1091,"_454":98,"_180":1031},{"_80":853,"_453":1020,"_454":31,"_180":1021},{"_80":1032,"_453":1091,"_454":31,"_180":1033},{"_80":1094,"_453":568,"_454":98,"_180":1095},{"_80":1092,"_453":1068,"_454":98,"_180":1093},{"_80":1026,"_453":1091,"_454":31,"_180":1028},{"_80":1089,"_453":568,"_454":31,"_180":1090},{"_80":853,"_453":1012,"_454":31,"_180":1021},{"_80":1017,"_453":1088,"_454":31,"_180":1019},{"_80":865,"_453":1018,"_454":31,"_180":1034},{"_80":867,"_453":1018,"_454":31,"_180":1085},{"_80":1032,"_453":1027,"_454":31,"_180":1033},{"_80":1086,"_453":666,"_454":31,"_180":1087},{"_80":867,"_453":692,"_454":31,"_180":1085},{"_80":1083,"_453":568,"_454":31,"_180":1084},{"_80":859,"_453":692,"_454":31,"_180":1078},{"_80":1055,"_453":671,"_454":98,"_180":181},{"_80":853,"_453":1082,"_454":31,"_180":1021},{"_80":1080,"_453":1020,"_454":98,"_180":1081},{"_80":1044,"_453":568,"_454":31,"_180":1035},{"_80":1017,"_453":568,"_454":31,"_180":1019},{"_80":1079,"_453":580,"_454":31,"_180":181},{"_80":859,"_453":568,"_454":31,"_180":1078},{"_80":867,"_453":1012,"_454":31,"_180":1077},{"_80":1075,"_453":675,"_454":98,"_180":1076},{"_80":1073,"_453":568,"_454":98,"_180":1074},{"_80":1072,"_453":568,"_454":31,"_180":181},{"_80":1070,"_453":1018,"_454":31,"_180":1071},{"_80":1069,"_453":580,"_454":31,"_180":181},{"_80":1060,"_453":1068,"_454":98,"_180":1062},{"_80":1066,"_453":671,"_454":31,"_180":1067},{"_80":1064,"_453":756,"_454":31,"_180":1065},{"_80":1063,"_453":574,"_454":31,"_180":181},{"_80":1060,"_453":1061,"_454":98,"_180":1062},{"_80":1058,"_453":1027,"_454":98,"_180":1059},{"_80":1056,"_453":568,"_454":31,"_180":1057},{"_80":1055,"_453":571,"_454":98,"_180":181},{"_80":1026,"_453":1030,"_454":31,"_180":1028},{"_80":858,"_453":1045,"_454":31,"_180":1035},{"_80":1053,"_453":571,"_454":31,"_180":1054},{"_80":1051,"_453":568,"_454":31,"_180":1052},{"_80":1049,"_453":568,"_454":31,"_180":1050},{"_80":1048,"_453":568,"_454":31,"_180":181},{"_80":890,"_453":571,"_454":31,"_180":1047},{"_80":1044,"_453":692,"_454":31,"_180":1035},{"_80":847,"_453":663,"_454":31,"_180":1046},{"_80":1044,"_453":1045,"_454":31,"_180":1035},{"_80":1043,"_453":580,"_454":31,"_180":181},{"_80":1042,"_453":568,"_454":31,"_180":1015},{"_80":1040,"_453":1027,"_454":31,"_180":1041},{"_80":853,"_453":1038,"_454":31,"_180":1039},{"_80":1037,"_453":580,"_454":31,"_180":181},{"_80":1036,"_453":568,"_454":31,"_180":181},{"_80":1029,"_453":1027,"_454":98,"_180":1031},{"_80":858,"_453":692,"_454":31,"_180":1035},{"_80":865,"_453":692,"_454":31,"_180":1034},{"_80":1032,"_453":1030,"_454":31,"_180":1033},{"_80":1029,"_453":1030,"_454":98,"_180":1031},{"_80":1026,"_453":1027,"_454":31,"_180":1028},{"_80":1024,"_453":1020,"_454":98,"_180":1025},{"_80":1022,"_453":678,"_454":31,"_180":1023},{"_80":854,"_453":1020,"_454":31,"_180":1021},{"_80":1017,"_453":1018,"_454":31,"_180":1019},{"_80":847,"_453":670,"_454":31,"_180":1016},{"_80":1014,"_453":568,"_454":31,"_180":1015},{"_80":865,"_453":1012,"_454":31,"_180":1013},{"_80":1010,"_453":568,"_454":31,"_180":1011},"LoginHints","Keeps users logged in across NVIDIA domains.","api.ngc.nvidia.com","This cookie is managed by AWS and is used for load balancing.","bm_so","This cookie name is used to load the site as quickly as possible and to recognize bots using Akamai.","The __cf_bm cookie is a cookie necessary to support Cloudflare Bot Management, currently in private beta. As part of our bot management service, this cookie helps manage incoming traffic that matches criteria associated with bots.\nThis is a CloudFoundry cookie","OptanonConsent","nvid.nvidia.com","This cookie is set by the cookie compliance solution from OneTrust. It stores information about the categories of cookies the site uses and whether visitors have given or withdrawn consent for the use of each category. This enables site owners to prevent cookies in each category from being set in the users browser, when consent is not given. The cookie has a normal lifespan of one year, so that returning visitors to the site will have their preferences remembered. It contains no information that can identify the site visitor.","events.nvidia.com","This cookie is associated with the Amazon Web Services Elastic Load Balancing functionality for routing client request on the server.Used to attribute commission to affiliates when you arrive at the website from an affiliate referral link. It is set when you click on one of our links and used to let the advertiser and us know the website from which you came.","sessionid","This is a very generic cookie name that may have different purposes on different sites, but generally it will be some kind of anonymous session identifier.","_csrf","This cookie is used to prevent Cross-Site Request Forgery (CSRF) attacks, ensuring that each request made to the server is from the user's browser and is intentional.","LSKey-c$CookieConsentPolicy","enterprise-support.nvidia.com","Used to apply end-user cookie consent preferences set by our client-side utility.Custom- Short Duration. Unique ID of an Enquiry- used to validate access to case, and load exisiting data.\n","renderCtx","academicgrants.nvidia.com","Used to store site parameters in the session for reuse across requests by a single client for functionality and performance reasons.","CookieConsentPolicy","Used to apply end-user cookie consent preferences set by our client-side utility.","This cookie is managed by AWS and is used for load balancing.Used to attribute commission to affiliates when you arrive at the website from an affiliate referral link. It is set when you click on one of our links and used to let the advertiser and us know the website from which you came.","This cookie is associated with Akamai and is used to differentiate between traffic from humans and bots.","nv_co","__uxq412__ses.88c1","learn.learn.nvidia.com","This cookie is associated with the Amazon Web Services Elastic Load Balancing functionality for routing client request on the server.","pctrk","Used to count page views by unauthenticated users against license usage.","bm_s","bpazaws52gukakzc__ctrl0_ctl36_uccaptcha","bm_sv","store.nvidia.com","The __cf_bm cookie is a cookie necessary to support Cloudflare Bot Management, currently in private beta. As part of our bot management service, this cookie helps manage incoming traffic that matches criteria associated with bots.\n","This cookie name is used by DataDog to determine if cookies are supported on the website.dd_cookie_test","AMP_TLDTEST_xxxxxxxx","SESSlsd","This cookie is likely to be a session identifier used to maintain user session state across page requests.","mboxEdgeCluster","This cookie name is associated with Adobe Target which is used to communicate the correct edge cluster to at.js so that Target profiles can stay in sync as users navigate across a site.","_rdt_em","This cookie name is associated with Reddit which is used to manage session state and user interactions.","megaNavCollapsed","_abck","The _abck cookie is associated with Akamai's Bot Manager, which is used to prevent fraud and abuse by safeguarding online interactions and detecting irregularities that could suggest malicious activity, such as bot-driven attacks or hacking attempts.","idccsrf","This cookie name is associated with Salesforce which is used to track CrossSiteRequestForgery validation for certain single sign-on (SSO) flows.","ApplicationGatewayAffinity","partners.nvidia.com","This cookie is used to maintain a user session by the server in a load balanced environment, to ensure the user has a consistent experience.","AMP_TEST_xxxxxxxx","bm_lso","This cookie name is associated with Akamai which is used to load the site as quickly as possible and to recognize bots.","SSESS###########","This cookie is likely to be a session cookie used to maintain an anonymized user session by the server in a secure environment, often seen in sites built with the Drupal content management system.","nvdeveloper.nvidia.com","__uxq412__id.88c1","OptanonAlertBoxClosed","This cookie is set by websites using certain versions of the cookie law compliance solution from OneTrust. It is set after visitors have seen a cookie information notice and in some cases only when they actively close the notice down. It enables the website not to show the message more than once to a user. The cookie has a normal lifespan of one year and contains no personal information.","QueueITAccepted-SDFrts345E-V3_{eventId}","TEST_AMCV_COOKIE_WRITE","This cookie is used to test whether cookies are enabled in the browser. It is a temporary cookie used to determine if the browser accepts cookies.","XSRF-TOKEN","This cookie is written to help with site security in preventing Cross-Site Request Forgery attacks.This cookie helps with site security in preventing Cross-Site Request Forgery attacks.","AWS ELB application load balancer","This cookie is associated with Akamai. It is used to analyse traffic to determine if it is automated traffic generated by bots or a human user.","bpazaws52gukakzc__ctrl0_ctl78_list_ctl09_systemcaptcha","PHPSESSID","PHP session cookie associated with embedded content from this domain.\n","base.manager.learn.nvidia.com","bm_ss","This cookie is used to help protect websites from automated bot traffic.","AWS ELB application load balancerUsed to attribute commission to affiliates when you arrive at the website from an affiliate referral link. It is set when you click on one of our links and used to let the advertiser and us know the website from which you came.","_devzone_session","This cookie is likely used to maintain a user session on the website by the server. It is common for session identifiers to be used to keep users logged in and track their interaction with the site during a session.","docs.ngc.nvidia.com","bm_sz","This cookie is used by Akamai to optimize site performance and security. It is typically associated with the delivery of web content and services.","vcalliance.nvidia.com","ASP.NET_SessionId","General purpose platform session cookie, used by sites written with Microsoft .NET based technologies. Usually used to maintain an anonymised user session by the server.","com.adobe.alloy.getTld","This cookie name is associated with the Adobe WebSDK which is used to determine the top-most domain that can accept cookies.","SESSls","This cookie is likely to be a session identifier, used to maintain a user's session state across page requests.","c_code","BIGipServerx","This cookie name is associated with the BIG-IP product suite from company F5. Usually associated with managing sessions on load balanced servers, to ensure user requests are routed consistently to the correct server. The common root is BIGipServer most commonly followed by a domain name, usually the one that it is hosted on, but not always.","This cookie is set by websites using certain versions of the cookie law compliance solution from OneTrust. It is set after visitors have seen a cookie information notice and in some cases only when they actively close the notice down. It enables the website not to show the message more than once to a user. The cookie has a one year lifespan and contains no personal information.","default","

NVIDIA uses cookies to improve your experience on our web site. We and our third-party partners also use cookies and other tools to collect and record information you provide as well as information about your interactions with our websites for performance improvement, analytics, and to assist in marketing efforts. By clicking \"Accept All\", you consent to our use of cookies and other tools as described in our Cookie Policy. You can manage your cookie settings by clicking on \"Manage Settings.\" By continuing to use this site or by clicking one of the buttons below, you agree to our Terms of Service (which contains important waivers). Please see our Privacy Policy for more information on our privacy practices.

","gpc","

We have detected the Global Privacy Control (GPC) signal and recorded your rejection of all optional cookies on this site for this browser. You can manage your cookie settings by clicking on \"Manage Settings\". Please see our Cookie Policy for more information. To opt out of non-cookie personal information \"sales\" / \"sharing\" for targeted advertising purposes, please visit the NVIDIA Preference Center. Please see our Privacy Policy for more information on our privacy practices.

","gpcLoggedIn","

We have detected the Global Privacy Control Signal (GPC) and have opted you out of all optional cookies on this browser. You can manage your cookie settings by clicking on \"Manage Settings\". Please see our Cookie Policy for more information. We have also opted you out of \"sharing\"/\"sales\" of personal information outside of cookies. You can manage these settings in the NVIDIA NVIDIA Preference Center. Please see our Privacy Policy for more information.

","gpcLoggedInOverridden","

We have detected the Global Privacy Control Signal (GPC) and have opted you out of all optional cookies on this browser. You can manage your cookie settings by clicking on \"Manage Settings\". Please see our Cookie Policy for more information. We have also opted you out of \"sharing\"/\"sales\" of personal information outside of cookies which overrides at least one of your previous settings. You can manage them in the NVIDIA Preference Center. Please see our Privacy Policy for more information.

",{"_792":18,"_695":55,"_603":55,"_424":55,"_414":55},"hasConsented","gpcApplied","showBanner","geolocation",{"_1116":1117,"_1118":1119},"country","US","state","OR","query","categories",[1139,1140,1141,1142,1143],"optionUrls",[1127,1128,1129,1130,1131,1132,1133,1134,1135,1136,1137,1138],"total",12,"/orgs/nim/deepseek-ai/containers/deepseek-r1/-","/orgs/nim/nvidia/containers/llama-3.1-nemotron-70b-instruct/-","/orgs/nvidia/-/containers/pytorch/-","/orgs/nvidia/omniverse/collections/kit/-","/orgs/nvidia/-/collections/deepstream_sdk/-","/orgs/nvidia/omniverse/collections/kit-appstreaming-collection/-","/orgs/nvidia/research/models/stylegan3/-","/orgs/nvidia/tao/models/peoplenet/-","/orgs/nvidia/tao/models/trafficcamnet/-","/orgs/nvidia/riva/resources/riva_quickstart/-","/orgs/nvidia/-/helm-charts/gpu-operator/-","/orgs/nvidia/-/helm-charts/k8s-nim-operator/-",{"_241":1215,"_1145":1216,"_1147":55},{"_241":1180,"_1145":1181,"_1147":16},{"_241":242,"_1145":1165,"_1147":1166},{"_241":1157,"_1145":1158,"_1147":1159},{"_241":1144,"_1145":1146,"_1147":143},"helm-charts","items",[1148,1149],"startIndex",{"_241":1144,"_248":1154,"_174":175,"_176":-7,"_243":1155,"_251":252,"_283":1156,"_1153":1137},{"_241":1144,"_248":1150,"_174":175,"_176":-7,"_243":1151,"_251":252,"_283":1152,"_1153":1138},"NVIDIA NIM Operator","k8s-nim-operator",[],"url","GPU Operator","gpu-operator",[],"resources",[1160],9,{"_241":1157,"_248":1161,"_174":175,"_176":1162,"_243":1163,"_251":252,"_283":1164,"_1153":1136},"Riva Skills Quick Start","riva","riva_quickstart",[],[1167,1168,1169],6,{"_241":242,"_248":1176,"_174":175,"_176":1177,"_243":1178,"_251":252,"_283":1179,"_1153":1133},{"_241":242,"_248":1173,"_174":175,"_176":177,"_243":1174,"_251":252,"_283":1175,"_1153":1134},{"_241":242,"_248":1170,"_174":175,"_176":177,"_243":1171,"_251":252,"_283":1172,"_1153":1135},"TrafficCamNet","trafficcamnet",[],"PeopleNet","peoplenet",[],"StyleGAN3 pretrained models","research","stylegan3",[],"collections",[1182,1183,1184],{"_241":1180,"_248":1212,"_174":175,"_176":1186,"_243":1213,"_251":252,"_283":1214,"_1153":1130},{"_241":1180,"_248":1209,"_174":175,"_176":-7,"_243":1210,"_251":252,"_283":1211,"_1153":1131},{"_241":1180,"_248":1185,"_174":175,"_176":1186,"_243":1187,"_251":252,"_283":1188,"_1153":1132},"Omniverse Kit App Streaming","omniverse","kit-appstreaming-collection",[1189,1190,1191],{"_80":1203,"_248":1204,"_245":1205,"_256":1206,"_1196":1207},{"_80":1186,"_248":1199,"_245":1200,"_256":1195,"_1196":1201},{"_80":1192,"_248":1193,"_245":1194,"_256":1195,"_1196":1197},"omniverse-dev","NVIDIA Developer Program","Join the Developer Program for access to free tools, support, and tech resources.","https://store-images.s-microsoft.com/image/apps.33936.6f1bcf42-55d0-45d1-b4b6-f08c68a50bbe.e91ca707-db64-49d6-9761-4781590a9685.7ba64fc5-7c44-4864-a5e6-7bde05126935","supportedFeatures",[1198],"UNIFIED_CATALOG","NVIDIA Omniverse","NVIDIA Omniverse is a platform of APIs, services, and software development kits (SDKs) that enable developers to build generative AI-enabled tools and applications.",[1198,1202],"SHAREABLE_TO_USER_GROUPS","nv-ai-enterprise","NVIDIA AI Enterprise","Accelerate your AI agent development","https://assets.ngc.nvidia.com/products/catalog/images/product-NVAIE.jpg",[1198,1202,1208],"NGC_CONSOLE","DeepStream SDK","deepstream_sdk",[],"Omniverse Kit (FB)","kit",[1189,1190,1191],"containers",[1217,1218,1219],{"_241":1215,"_248":1227,"_174":1224,"_176":1228,"_243":1229,"_251":252,"_283":1230,"_1153":1127},{"_241":1215,"_248":1223,"_174":1224,"_176":175,"_243":1225,"_251":252,"_283":1226,"_1153":1128},{"_241":1215,"_248":1220,"_174":175,"_176":-7,"_243":1221,"_251":252,"_283":1222,"_1153":1129},"PyTorch","pytorch",[],"Llama-3.1-Nemotron-70B-Instruct","nim","llama-3.1-nemotron-70b-instruct",[],"DeepSeek-R1","deepseek-ai","deepseek-r1",[],"ownerOrgs",[],"memberOrgs",[],"externalOrgs",[],"apiKey","2b44f86fea50e90ffd2db29ee2028419","consentGranted","applicationId","4c857e4f-5ffa-45fa-b75e-af773c22f0ad","clientToken","pubb6927db33dae60ba5031a186cde0242a","site","datadoghq.com","service","catalog-ui","env","sessionSampleRate",100,"sessionReplaySampleRate","compressIntakeRequests","storeContextsAcrossPages","defaultPrivacyLevel","mask-user-input","trackViewsManually","trackingConsent","not-granted",{"_1260":1261,"_1262":361,"_1263":31,"_1264":1265,"_1145":1266,"_1267":1268,"_1269":1261,"_1270":31,"_1271":31,"_1272":31,"_1273":1274},"appName","ngc-catalog","appOrigin","hideOnPrivateOrg","icon","ObjectsModules2",[1275,1276,1277,1278,1279,1280,1281],"menuIcon","editor-swatches","productName","requiresAllScopes","requiresLogin","requiresOrgLevelScopesOnly","title","Catalog",{"_1273":1303,"_47":1304,"_1267":1305},{"_1296":1297,"_1273":1298,"_47":1299,"_1269":1203},{"_1273":1294,"_47":1295,"_1267":1268},{"_1273":1291,"_47":1292,"_1267":1293},{"_1273":1288,"_47":1289,"_1267":1290},{"_1273":1285,"_47":1286,"_1267":1287},{"_1273":1282,"_47":1283,"_1267":1284},"Resources","/resources","misc-library","Models","/models","misc-neural-network","Helm Charts","/helm-charts","misc-ship-wheel","Containers","/containers","3d-cube","Collections","/collections","access",[1300],"Enterprise Catalog","/enterprise",[1301,1302],"NVAIE","READ","Explore Catalog","/","common-home",{"_124":125,"_126":127,"_128":129,"_130":131,"_132":133,"_134":135},{"_123":31}]

TAO Sparse4D Model Card

Model Overview

Description:

License/Terms of Use:

Deployment Geography:

Use Case:

Release Date:

References

Model Architecture

Input:

Output:

Software Integration:

Model versions (Latest):

Training, Testing, and Evaluation Datasets

Dataset Overview

Data Format & Preprocessing

Raw Data (MTMC Tracking 2025/AI City Challenge Format)

Pickle File Structure (OVPKL format)

metadata key

infos key

cams Structure

Public Datasets

Private Datasets

Self-Sourced Synthetic Data

Training Data

Testing and Evaluation Datasets

Methodology and KPI

ResNet50 backbone (v2.2)

ResNet101 backbone (v2.2)

Real-time Inference Performance

Changelog

v2.2

v2.1

v2.0

v1.1

v1.0

How to use this model

Inference & Deployment:

Ethical Considerations

`metadata` key

`infos` key

`cams` Structure

Name	Size	Updated	Actions
_ov_kmeans900_v0.5.npy	77.47 KB	February 9, 2026 UTC	curl wget Download
experiment.yaml	6.86 KB	February 9, 2026 UTC	curl wget Download
sparse4d_tracking_aic25v0.5_cosmosv0.1_warehousev3_classes_iter_46400_v2.0.pth	767.19 MB	February 9, 2026 UTC	curl wget Download