face extraction

Installed on pia inside /userdata/bdichter/envs/py35

Installation

I recommend using a conda envoronment. ffmpeg and dlib are required. If on mac, install ffmpeg by running brew install ffmpeg. Installing dlib on a mac is a bit of an ordeal. This post should help. It also uses opencv3, however I was only able to get this to work in python 3.5, so I used conda virtual environment. Putting it all together (assuming anaconda is properly installed):

brew install ffmpeg
conda create -n face_extraction_env python=3.5 pillow numpy
source activate face_extraction_env
conda install -c menpo opencv3
pip install tqdm
pip install face_recognition

Usage

from face_extraction import detect_faces_video
detect_faces_video('video_in.mp4', 'video_out.avi')

OpenCV is pretty picky about the form of the input video, so you may have to use ffmpeg to change the video codec. I have had the most luck with .mp4 files with an mpeg4 vcodec. The recording setup uses h256, which doesn't work right away. Here is an example convertion:

ffmpeg -i SN6.mp4 -vcodec mpeg4 SN6_v2.mp4

Ideas

• Implement in C++
• split video up with ffmpeg, run on different cores, then recombine them