README.md

Video analysis demo

Requirements

Python

• openCV with video support (Instruction)
• tensorflow (Instruction)
• devicehive-python-webconfig (for daemon.py)
• pafy (for direct video links)
• youtube-dl (for direct video links from youtube)

All requirements are listed in requirements file, except openCV (because of video support described here, here and here)

So (almost) all required packages can be installed by pip

Update: Since this release opencv-python ships with FFmpeg, V4L and V4L2. So i have added it to requirements.

pip install -r requirements.txt

Models data

• download and extract data.tar.gz to source folder

Running

Local demo

To evaluate video file run

python eval.py --video="/path_to_video_file/"

Also video url can be passed (tested for youtube)

python eval.py --video="youtube_url"

If --video is not provided video device "0" (usually it's web cam) will be used by default.

Press q to close program.
Press s to save currents frame to file.

Run

python eval.py --help

for more info about available arguments.

Web based demo

There is demo with web interface and devicehive integration. It will capture video stream from your web cam, evaluate it and send predictions to devicehive. To use it run

python daemon.py

and go to http://127.0.0.1:8000 to configure devicehive connection. Video stream is available on http://127.0.0.1:8000/events/