Ab-Esc is a database for SARS-CoV-2 immune escape mutations. We have developed a machine learning model to classify SARS-CoV-2 RBD point mutations as leading to high or low antibody escape using change in binding affinity upon mutation. For developing the model we used experimental escape fraction data available in literature, known as Dataset 1 in our study. Based on the model, we made the predictions on a blind dataset of 83 SARS-CoV-2 neutralizing antibodies (Dataset 2). Our data contains information about the escape fraction, interface mutations, RBD site, change in binding affinity upon mutation from different methods, etc.