Sign_language_detector

This is a sign language detector,based on american sign language.Made using tensorflow based on Sequential model,having used Convlution layers and Maxpooling on the top of a DNN.The Optimizer used is rmsprop and the loss function used is sparse categorical cross entropy.

Dependencies-

-Cv2

-Tensorflow(2.0)

-Numpy

-Matplotlib

-PIL

How to run the script

There is a trained model already present in the files,you can also try different models in the jupyter notebook(untitled 1) to get started,and make model better.

Current accuracy is about 90% on validation data set.

Data set is already given in the repo for use .

Run the final.py file

How to get the best results

Try keeping a simple plane background,Keep your hand in the green box,Try avoiding excessive light or Shadow. Try keeping your face out of the box

Press q to click the image you want to process.