ContainersNVIDIADeepSeek-R1NVIDIALlama-3.1-Nemotron-70B-InstructNVIDIAPyTorch
CollectionsNVIDIAOmniverse Kit (FB)NVIDIADeepStream SDKNVIDIAOmniverse Kit App Streaming
ModelsNVIDIAStyleGAN3 pretrained modelsNVIDIAPeopleNetNVIDIATrafficCamNet
ResourcesNVIDIARiva Skills Quick Start
Helm ChartsNVIDIAGPU OperatorNVIDIANVIDIA NIM Operator
NVIDIA
NVIDIA
Sparse4D (ResNet-101)
Model
NVIDIA
NVIDIA
Sparse4D (ResNet-101)

TAO Pretrained Sparse4D with ResNet-101 Backbone

This model is backed by NVIDIA's Plus Plus (++) Promise
to learn more about the quality of the datasets used to train this model.
FieldResponse
Intended Application(s) & Domain(s):3D Multi-Camera Detection & Tracking of Objects.
Model Type:Resnet101 Backbone + Transformer
Intended Users:This model is intended for developers working in industrial applications.
Output:Bounding box information & object IDs in 3D coordinates
Describe how the model works:Input images from multiple cameras are encoded & passed onto decoder layers for regression & classification tasks.
Name the adversely impacted groups this has been tested to deliver comparable outcomes:Not Applicable.
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.
Verified to have met prescribed NVIDIA standards:Yes
Performance Metrics:Average Precision (AP) & Mean Average Precision (mAP)
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.
Licensing:GOVERNING TERMS: Use of this model is governed by the NVIDIA Open Model License.