Emotion Recognition

It is the repository for the code presented on the article Building and validation of a set of facial expression images to detect emotions: a transcultural study by Julian Tejada, Raquel Meister Ko. Freitag ,Bruno Felipe Marques Pinheiro, Paloma Batista Cardoso, Victor Rene Andrade Souza, Lucas Santos Silva. The code is divided in three scripts, which represent the three necessary steps to process a video:

1. Extract the frames from the video
2. Recognize faces o each frame and cut them into a png B&W image
3. Classify the faces B&W image using some of the images dataset.

To extract the frames

python Script_CutVideos.py <name of the video file (mp4)> <Images folder path> <start frame> <stop frame>

To cut the faces of the frames

python ScriptExtractingFaces.py <Source images path> <Destination path (Faces B&W images)>

To classify

python Script_EmotionClassification.py <Source images path> <dataset train (npy)> <dataset labels (npy)>

The second and third scripts need the landkmarks descriptions find in the next files

1. From the dlib project the Shape predictor 69 face landmarks
2. From OpenCV project the haarcascade_frontalface_default.xml, haarcascade_frontalface_alt.xml, haarcascade_frontalface_alt2.xml, haarcascade_frontalface_alt2.xml

Datasets

Two different datasets of landmarks extracted from images of face expressions (six basic emotion plus mockery) from a transcultural sample of participants (Brazilian and Colombian subjects). The dataset are in two different files, one whit the landmarks (Train_Brazil.npy and Train_Colombia.npy) and other with the class (Train_Brazil_labels.npy and Train_Colombia_labels.npy) where the class order is: 0.disgust, 1.happiness, 2.surprise, 3.anger, 4.sadness, 5.neutral, 6.fear, 7.mockery.