As the traditional computer vision solutions were not able to recognize all the printed characters in a scanned cheque, this solution based on deep learning provided an acceptable level of results - More than 80% as opposed the traditional open cv based approach which yielded 60%. This involved annotating and training close to 10000 cheques.I was instrumental in designing and developing the entire solution.