People recognition through HTTP video signal in docker.
This project makes use of the SSD Mobilenet model trained in COCO dataset to perform people detection from the video signal received by an HTTP signal. It then publishes a UI delivered by an HTTP Server that shows the amount of people detected in that frame and that frame with the detection boxes and the likelihoods.
Build the image through Docker Compose:
cd project_route
docker-compose -d --build .
The image will take 2.67 GB, so it may take several minutes to get ready.
Prepare the signal-sender machine to deliver the video signal through HTTP (for example with VLC). Then change the required variables stored in the .env file, which are:
- VIDEO_IP: HTTP address sended for it's processing.
- IMAGE_REFRESH: Time interval (in seconds) taken by the algorithm to process the next frame. 0.5 seconds by default.
- DOCKER_PORT: Internal docker port from where launch
ui.html. 8085 by default.- UI_PORT: Port used by the processing machine to deliver the web UI. 8085 by default.
Finally, type docker-compose up, open a web browser and go to YOUR_IP:UI_PORT to check the result. Press Ctrl + C to stop execution.