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
SyntheticaDETR
Model
NVIDIA
NVIDIA
SyntheticaDETR

SytheticaDETR is a real-time object detection model based on a transformer architecture trained entirely in simulation and works on real images zero-shot.

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):Learning to detect objects in a scene with prediction confidence suitable for robotics and autonomous machines.
Model Type:Object detection model.
Intended Users:Developers building and/or customizing robotics applications.
Output:Bounding box coordinates and confidence.
Describe how the model works:Model detects objects in a scene.
Technical Limitations:SyntheticaDETR is not able to accurately detect objects outside the class of objects it was trained on.
Performance Metrics:Mean Average Precision (mAP)
Potential Known Risks:N/A
Licensing:https://developer.download.nvidia.com/licenses/tao_toolkit_21-08_models_eula.pdf