Convolutional Neural Networks for Facial Expression Recognition

In the context of this study, convolutional neural networks CNNs have been developed with the objective of identifying facial expressions. The primary aim of this study project is to categorize each facial image into one of the seven distinct categories of facial expressions under investigation. The training of Convolutional Neural Network CNN models with different levels of depth included the use of grayscale photographs sourced from the Kaggle website 1 . By using Torch 2 , we successfully developed our models and used the computational capabilities of Graphics Processing Units GPUs to enhance the efficiency of the training procedure. In addition, we used a hybrid feature method alongside the networks that were operating on raw pixel data. By integrating raw pixel data with Histogram of Oriented Gradients HOG characteristics, we were able to train a distinctive CNN model 3 . We used several techniques, including dropout and batch normalization, along with L2 regularization, to mitigate the occurrence of overfitting in the models. Cross validation was used to determine the optimal hyper parameters, and the performance of the generated models was assessed by examining their individual training histories. Furthermore, we provide a visual representation of the several layers inside a network to illustrate the attributes of a facial feature that may be acquired using Convolutional Neural Network CNN models. Rashmi Yadav | Mr. Ghanshyam Sahu | Lalitkumar P Bhaiya "Convolutional Neural Networks for Facial Expression Recognition" Published in International Journal of Trend in Scientific Research and Development (ijtsrd), ISSN: 2456-6470, Volume-8 | Issue-3 , June 2024, URL: https://www.ijtsrd.com/papers/ijtsrd64783.pdf Paper Url: https://www.ijtsrd.com/computer-science/artificial-intelligence/64783/convolutional-neural-networks-for-facial-expression-recognition/rashmi-yadav