This repository showcases the use case of image classification on SIGN dataset with 6 classes. The dataset contains hand signs. The model used has CONV2D -> RELU -> MAXPOOL -> CONV2D -> RELU -> MAXPOOL -> FLATTEN -> FULLYCONNECTED layers. Implementation is done in pytorch. The model achieves an accuracy of 99% on train data and 91% on the test data. I had implemented a flask API for this application which takes an image as input and outputs the image prediction.
Below image shows how an image is uploaded through the Postman(client), and the format of response returned by the server.