This project implements a Neural Network(with TensorFlow 1.0 and Python 3) to interpret sign language. The dataset is a subset of the SIGNS database.
The dataset is divided in the following way:
TRAINING SET - 1080 photos (64 by 64 pixels) representing numbers from 0 - 5
TEST SET - 120 photos (64 by 64 pixels) representing numbers from 0 - 5
Training Accuracy achieved - 100%
Testing Accuracy achieved - 78%
NB. the difference in training and testing accuracy suggests overfitting.
The model uses Adam Optimizer to optimise the cost function.
Code contributed by Sammya Majumdar