sumansid/sign-language-recognition

A deep learning model that recognizes hand gestures for alphabets. Trained using tensorflow, with activation function : RELU and Softmax (for multi-class classification).

Sign-Language-Recognition

A deep learning model that recognizes sign language alphabets.

Number of Hidden Layers : 3

A deep neural network is used to gather train on more features. Currently, we have 25 classes, thus a deep NN is needed to distinguish each alphabet from the other and achieve higher accuracy.

Number of training examples = 27455 Number of test examples = 7172

X_train shape: (784, 27455), Y_train shape: (25, 27455), X_test shape: (784, 7172), Y_test shape: (25, 7172)

Number of channels in each image : 1 (Grayscale), Size of each image vector = 28 x 28 x 1 = 784

Tools and Deep Learning Framework : Tensorflow Version 2.0, Numpy and Pandas

Model : Linear -> Relu -> Linear -> Relu -> Linear -> Softmax

Hand Gestures :

Dataset : https://www.kaggle.com/datamunge/sign-language-mnist

Contributing

Pull requests are welcome.