Green Screen when launching yolox_ros_py

[JonathanNash21]

Hello,

When I run yolox_ros_py on my Jetson Nano, I encounter a green screen like in the screenshot. This happens when using yolox_nano_torch for both the cpu and gpu versions - the Docker container I'm using only has PyTorch, so I can't run the other options. I've checked running a GStreamer application, and that works both in the native environment and in the Docker container I'm running yolox_ros in - I think the issue might be with v4l2 or CvBridge, but I'm not entirely sure. Is there an easy way to apply GStreamer instead?

I've tried using the Dockerfile for Jetson Nano found in the yolox_ros_cpp folder, but the build fails at the 19th and 21st build commands (installing the onnxoptimizer from git and installing YOLOX from git) - if you have this image hosted on Dockerhub, I should be able to test and see if that will work by just downloading the built image.

[JonathanNash21]

I will add that I edited the dockerfile to be pull from dustynv/ros:galactic-ros-base-l4t-r32.7.1, I wouldn't think this would cause issues with building though

[Ar-Ray-code]

I think the issue might be with v4l2 or CvBridge, but I'm not entirely sure. Is there an easy way to apply GStreamer instead?

If you think v4l2-camera is the cause, then I don't think this issue comes from Jetpack. rqt_graph should see if it shows the screen.

Checking that the YOLOX demo.py works correctly in the container would be a good way to get some hints. Please give it a try.

[JonathanNash21]

Sorry, I should've gotten back to this earlier - I fixed the issue by changing the pipeline to use GStreamer through a combination of using https://github.com/BrettRD/ros-gst-bridge instead of v4l2-camera as the webcam node in the launch file (not sure if this does anything important due to the next thing) and then using a gstreamer pipeline via OpenCV to open the camera in yolox_ros.py, which I convert to a ROS message via the CvBridge to be able to publish the video frames.