This repo demos the frameowork, Deepstream, that it will consume a video or a RTSP and do the object detection, then output the result as a RTSP.
Because everything is running on AWS, forwarding screen will be too annoying. Here we use ffmpeg to record the RTSP stream. Check dockerfile for more details.
- See instructions
- Follow https://stackoverflow.com/a/61737404 first
- When it builds, it will automatically run a demo that consumes a video and store the output RTSP as videos with two different methods.
- binary: executed by a compiled binary execution app with a config file.
- python: executed by a python script.
- Run
docker build -t neilvaltec/demo_deepstream:tag .
- Check the dockerfile for more details.
- You can check the deepstream output result during the image build under
/root/output/
- Check here to see how to run a container as a input RTSP.
-
docker run -it --rm --gpus all --network:container:<container_id> neilvaltec/demo_deepstream:tag bash
-
col: method
-
row: input source
video RTSP deepstream-app V V python script V TODO