Face Emotion Recognition from Images

This Project has 2 sub projects using 2 different models to detect emotions from images.

DeepFace

The Code under \emotion_detection_deepface uses the DeepFace library to detect emotions from images. The code is based on the DeepFace Library

FER 2013 based model

The Code under \emotion_detection_fer2013 uses the FER 2013 dataset to train a model to detect emotions from images. The code is based on the FER 2013 dataset The pre-trained model is saved in the \emotion_detection_fer2013\src folder. The model is trained for 50 epochs and has an accuracy of 63.2% on the test data.

To run the code, follow the steps below:

1. Copy all your images to a folder

2. Go to face_emotion_detection.py file,

3. Change the last line that calls load_and_analyze_all_faces(<images path>)

4. Replace the <images path> with the full path of where the images are located

5. Run the code and wait for the results

6. The results will be saved in an HTML file in the same folder as the images

7. The HTML file will have the image links and the Deepface analysis JSON + emotion detected from each image in 3 columns

8. The load_and_analyze_all_faces function calls the predict_emotion(image_path) method in emotion_predictor_fer2013.py file which is the main file that does the emotion detection based on fer2013 model