Using Convolutional Neural Network to predict the letter being signed from an image of a hand. The data was self-gathered from a group fo 40 students with 10 letters and 3 orientations of each letter. The Data was cleaned then pre-processed with normalization and finally balanced to ensure equal representation in both training and validation sets. The network is a CNN with 4 layers and Batch Normalization with an Adam optimizer and uses Cross-Entropy loss to achieve a validation accuracy of 84.6%
