Not able to run PeopleSegNet(MaskRCNN) NVIDIA TAO(.etlt) model with deep stream

[kotreshrr17]

Hi, I was trying to run PeopleSegNet(MaskRCNN) TAO(.etlt) model in deepstream. config file which is provided here(deepstream_tao_apps/configs/peopleSegNet_tao at master · NVIDIA-AI-IOT/deepstream_tao_apps · GitHub) and have build and replaced “libnvinfer_plugin.so*” as per given in this doc(deepstream_tao_apps/TRT-OSS/x86 at master · NVIDIA-AI-IOT/deepstream_tao_apps · GitHub). I was unable to run the model, please find the below screenshot for the error message. Could you please help us resolve this issue.

DeepStream config file: deepstream_tao_apps/pgie_peopleSegNet_tao_config.txt at master · NVIDIA-AI-IOT/deepstream_tao_apps · GitHub
Model: PeopleSegNet TAO model which is downloaded from https://api.ngc.nvidia.com/v2/models/nvidia/tao/peoplesegnet/versions/deployable_v1.0/zip.
Labels file : deepstream_tao_apps/peopleSegNet_labels.txt at master · NVIDIA-AI-IOT/deepstream_tao_apps · GitHub
Hardware Platofrm: Tesla T4
DeepStream Version: 6.0
TensorRT Version : 8.0.1
CUDA Version: 11.3

Thanks

[Morganh]

Can you try another config file deepstream_tao_apps/pgie_peopleSegNetv2_tao_config.txt at master · NVIDIA-AI-IOT/deepstream_tao_apps · GitHub ?

[kotreshrr17]

Hey Morganh, Thanks for the suggestion, I was able to create trt engine file using that v2 config and its model, but I am not able to parse the segmentation mask output, which was suppose to be in pyds.NvDsObjectMeta.mask_params, I was trying to access it, I am not able to do so, could you please help us on how to parse the output of the model and how do I get the mask info in the ObjectMeta. please find the below error message.

Thanks

[Morganh]

For postprocessing of maskrcnn, you can refer to tao-toolkit-triton-apps/peoplesegnet_postprocessor.py at main · NVIDIA-AI-IOT/tao-toolkit-triton-apps · GitHub

[kotreshrr17]

Thanks