Video-Expression-Recognition

Python script to recognise emotion of all facial images in a video. Given a video, the script will recognise all facial images and associate each face with an expression(anger, disgust, fear, sadness, surprise, joy and neutral).

Input-Output

Input

Video with/without faces

Output

For a chosen unique cluster(face), graph of how its facial expression/emotion is changing through time.

Requirements

• Keras=1.2.2
• Tensorflow=1.2.1
• Dlib=19.4.99
• OpenCV
• ffmpeg

The code is tested in a container built from Ubuntu 14.04 CPU docker image downloaded from floyd-hub(link given below).

Demo

1. Download necessary weight files from weight-files
2. Extract all files downloaded from the above directory to the directory containing demo.py
3. Call demo.py by running "python demo.py -v /full/path/to/video"
4. Follow runtime instructions

For a quick testing, use the video provided in example_videos directory.

References

• Docker Image
• Several ideas from ICMI2015-ChaZhang are used to create the expression recognition model.
• This code is built on top of Facial-Expression-Recognition