AlexeyAB / darknet

YOLOv4 / Scaled-YOLOv4 / YOLO - Neural Networks for Object Detection (Windows and Linux version of Darknet )
http://pjreddie.com/darknet/
Other
21.76k stars 7.96k forks source link

Publish Detections to WebRTC Server #1683

Open ashuezy opened 6 years ago

ashuezy commented 6 years ago

Sending opencv frame to WebRTC server looks like a good idea so that on a network the detections can be viewed by many in their browser. Can you suggest how to go about sending frames to WebRTC server ?

AlexeyAB commented 6 years ago

I didn't work with WebRTC. Only if your additional soft can get mjpeg http-stream from IP (computer with Yolo) and broadcast it to many users - then you can try to use such command:

./darknet detector demo data/voc.data cfg/yolov2-voc.cfg yolo-voc.weights test.mp4 -i 0 -http_port 8090 -dont_show

Then read http-stream by using URL: http://ip-address:8090

ashuezy commented 6 years ago

@AlexeyAB What if you want to broadcast N different CCTV cameras RTSP feeds ? Running each instance on a different port doesn't seem practical on a public server with N number of ports open.

I would like to play multiple streams on single port with different url.

ashuezy commented 6 years ago

@AlexeyAB

I have started this repo to address this issue:

https://github.com/ashuezy/mjpeg_server

There are few issues left to resolve but will post when it is working.