Hand gestures which are a form of non-verbal communication are a strong way to communicate with a hearing-impaired community. It helps to establish human-computer communication. The purpose of this project is to help the dumb and deaf people for ease in communication as this project will help us to understand their sign language to some extent as both hands are used for performing any gesture and this will eventually help them to communicate with others. Image classification is done using neural networks. The task is accomplished using Machine learning and Image classification. InceptionV2 model with feature vector is used to train the dataset of around 10000 images in total and 300 images of each sign which gives the optimum accuracy.