Welcome to the Facial Expression Recognition project! This project aims to develop a deep learning model for accurately classifying facial expressions and estimating arousal and valence values from images. The model is based on two different CNN architectures, ResNet and EfficientNetB0, which have been fine-tuned and trained on a large dataset of images and corresponding label files.
For this project we have used a private dataset, you can use your own dataset with specified annotations for class categories and arousal and valance values.
In this project, we utilized two popular CNN architectures as baselines for our model: ResNet and EfficientNetB0. ResNet is a deep residual neural network that has shown excellent performance in image classification tasks. It uses skip connections to allow information to pass through layers without being altered, thus enabling the network to learn more complex features. EfficientNetB0, on the other hand, is a lightweight and scalable CNN architecture that has been optimized for both accuracy and computational efficiency. It uses a combination of convolutional layers, squeeze-and-excitation blocks, and depthwise separable convolutions to extract features from images while minimizing the number of parameters and computations required. Both ResNet and EfficientNetB0 have been widely used in various computer vision applications, and we chose them as our baselines due to their proven performance in image classification tasks.
Israr Ahmed iahmed.msds22seecs@seecs.edu.pk