Open Access. Powered by Scholars. Published by Universities.®
Articles 1 - 2 of 2
Full-Text Articles in Engineering
Performance Comparison Of Scanner And Camera-Acquired Data For Bangla Offline Handwriting Recognition, Nishatul Majid, Elisa H. Barney Smith
Performance Comparison Of Scanner And Camera-Acquired Data For Bangla Offline Handwriting Recognition, Nishatul Majid, Elisa H. Barney Smith
Electrical and Computer Engineering Faculty Publications and Presentations
This paper presents a comparison of offline handwriting recognition performance between two different image acquisition devices: scanner and cell-phone camera. Whereas a flat-bed scanner offers higher quality distortion free imaging, a cell-phone camera trumps on the convenience and ease of use. The aim of this research is to quantify how the extra quality obtained from a scanner impacts the offline handwriting recognition. This was evaluated with two classification framework: a segmentation-free offline Bangla handwriting transcription with sequential detection of characters/diacritics and a Bangla handwritten digit recognizer with an SVM classifier. The Boise State Bangla Handwriting dataset is used for the …
Segmentation-Free Bangla Offline Handwriting Recognition Using Sequential Detection Of Characters And Diacritics With A Faster R-Cnn, Nishatul Majid, Elisa H. Barney Smith
Segmentation-Free Bangla Offline Handwriting Recognition Using Sequential Detection Of Characters And Diacritics With A Faster R-Cnn, Nishatul Majid, Elisa H. Barney Smith
Electrical and Computer Engineering Faculty Publications and Presentations
This paper presents an offline handwriting recognition system for Bangla script using sequential detection of characters and diacritics with a Faster R-CNN. This is an entirely segmentation-free approach where the characters and associated diacritics are detected separately with different networks named C-Net and D-Net. Both of these networks were prepared with transfer learning from VGG-16. The essay scripts from the Boise State Bangla Handwriting Dataset along with standard data augmentation techniques were used for training and testing. The F1 scores for the C-Net and D-Net networks are 89.6% and 93.2% respectively. Afterwards, both of these detection modules were fused into …