Title of Submission
Electrical and Computer Engineering, PhD
Major Advisor Name
Elisa Barney Smith
Type of Submission
This presents an offline handwriting recognizer for Bangla script. In spite of being a major script, very little progress has been made in this field for Bangla. Here, we present a handwriting recognition unit with sequential detection of characters/diacritics. A faster R-CNN was used to spot the graphemes from word images and the results were merged to form a transcription. Transfer learning and data augmentation techniques were applied to increase the speed and accuracy of the process. We achieved a WER and CER of 21.5% and 8.9% respectively, which is the first reported transcription result for Bangla script.