Research & Projects
1
Evaluating the Security of CAPTCHAs utilized on Bangladeshi Websites
In this research, our work is about determining the vulnerability of CAPTCHAs embedded in Bangladeshi websites by building an Evaluation Framework of how secure the CAPTCHAs are and a generalized attack method to break all types of text-CAPTCHAs. Then develop a secure Bengali CAPTCHA mechanism.
2
A Secure and Usable CAPTCHA for Bengali Language
In this research, We have designed a secure Bengali CAPTCHA mechanism that is secure from all the current CAPTCHA attacks. We have also done usability study of the CAPTCHA.
3
POS tagging of Code-Mixed Language(BN_EN)
In this research, We have build a parts of speech (POS) tagger for Bengali & English mixed sentences. We have collected Code-Mixed Bengali English Sentences from social media platform. Seperated the Bengali and English part using a Bi-LSTM language identifier, then used separate POS tagger for Bengali and English Language, then combine it by mapping into universal POS tagset.
4
Facial Expression Recognition using CNN and LSTM
The purpose of this research is to make a supervised model for recognizing Facial Expressions of human faces using two models CNN and LSTM. Our approach to this problem was quite new and different. We did train the model in such a way that the emotion’s expression is not just one single image, but a continuity of a sequence of images of that emotion. For datasets, we modified the FER2013 and CK+ datasets in such a way of augmentation that it produces a sequence of images, then make a sequence out of it