Only one class detected in dashcamnet

Please provide complete information as applicable to your setup.

• Hardware Platform (Jetson / GPU) :- GPU
• DeepStream Version :- 9.0
• TensorRT Version :- 10.14.1.48
• NVIDIA GPU Driver Version (valid for GPU only) :- 535
• Requirement details :- nvcr.io/nvidia/deepstream:9.0-triton-multiarch
• Model :- dashcamnet_pruned_onnx_v1.0.5

I am using the Docker image nvcr.io/nvidia/deepstream:9.0-triton-multiarch and have installed the DeepStream Python bindings using ./user_deepstream_python_apps_install.sh --build-bindings.

I am performing inference using the DashCamNet pruned model downloaded from NGC with the Python deepstream-test3 application. The application runs successfully, but I only observe detections for the Person class. The other expected classes (Car, Road Sign, and Bicycle) are not being detected. No errors or warnings are reported during inference.

Refrence documents are attached

pgie_label.zip (1.7 KB)

This model was released in 2024 and is incompatible with current CUDA, TensorRT, and DeepStream versions; furthermore, it offers functionality identical to /opt/nvidia/deepstream/deepstream/samples/models/Primary_Detector/resnet18_trafficcamnet_pruned.onnx. What is your objective? There might be a better way of doing things.