This project brings together learning and computer vision to help deaf and hearing people communicate better using sign language. We propose a system to recognise American Sign Language (ASL) using a built Convolutional Neural Network (CNN). The model was trained from scratch on a dataset of 5,583 images of 27 hand signs (the alphabet and a blank sign).To make the model better, we resized images to 64x64 pixels. Used various data augmentation techniques. This system can recognise hand gestures in real time using a webcam and achieved an accuracy of 87.33% on test data. Such a system can be very useful for real-time translation and making computers more accessible.