Implementation of Genetic Algorithm for finding the best subset of features for emotion detection from an image using the CK+ dataset and Openface toolkit.
- This model was trained on The Extended Cohn-Kanade Dataset (CK+) which is a complete dataset for action unit and emotion-specified expression
- The images from the dataset were then processed using the OpenFace toolkit to obtain the final dataset
- Manually included labels to the dataset in order to train the model for the respective emotion
- Random initialization of population chromosomes
- Fitness for the individuals (feature subset) was defined using the mean accuracy for classifying emotions
- Selection methods:
- Roulette-wheel selection
- Rank-based selection
- Tournament selection
- Crossover methods:
- k-point crossover
- Uniform crossover
- Mutation methods:
- Bit-swap mutation
- Bit-flip mutation
- Population update methods:
- Generational update
- Weak-parents update