Error when running Vision AI Movenet container: Could not configure supporting library

AI & Data Science Computer Vision & Image Processing
1

Hello,
I’m trying to use NVIDIA’s Vision AI Movenet container (nvcr.io/nvidia/ace/vision-ai-movenet:0.1.82) to run inference on local images or videos.

However, when I run the following command, I get an error: Could not configure supporting library, and the container does not function correctly.

sudo docker run --gpus all -it --rm \
  -p 8000:8000 \
  nvcr.io/nvidia/ace/vision-ai-movenet:0.1.82

Partial error log:

Error from ...: Could not configure supporting library.
unable to connect to broker library
Failed to set GStreamer pipeline to PLAYING
...

Environment:

  • Cloud platform: Azure VM (Standard NC6s v3, Tesla V100)
  • OS: Ubuntu 22.04
  • NVIDIA Driver: 565.57.01
  • CUDA Version: 12.6 (installed) / 12.7 (reported by nvidia-smi)
  • Docker: 20.10.x

What I’ve tried:
✅ Verified that nvcc --version confirms CUDA 12.6 is installed
✅ Checked nvidia-smi to confirm the GPU and driver are working correctly
✅ Retrieved CUDA version information from /usr/local/cuda-12.6/version.json

Questions:

  1. What is the correct way to use Movenet for inference on local images and videos?
  2. What could be causing the Could not configure supporting library error, and how can I fix it?
  3. How can I verify that the container is correctly utilizing the GPU at runtime?

I am new to DeepStream and Movenet, and I’ve been checking the official documentation, but I am struggling to figure this out.
Any advice or guidance would be greatly appreciated! Thank you in advance.