🤏Deep Learning based American Sign Language Recognition
✨Received extra credit for this project✨
- Developed three classes of Artificial Neural Network (ANN) models using the PyTorch framework to translate American Sign Language (Kaggle dataset) into a text-based representation.
- First class was a single generic Multilayer Perceptron (MLP) network.
- Second class was two untrained CNNs (Convolutional Neural Networks)- a baseline CNN and a custom CNN (with Adam optimizer and L2 regularization).
- Third class was used to leverage Transfer Learning with two pre-trained CNNs- GoogLeNet and ResNet18 networks.
- All five models were tested using images captured in real-time and achieved an accuracy of above 90%.
✦ Code: 
✦ Medium blog: 
✦ Report: 