Sujit Debnath

I am a graduate student in Artificial Intelligence at Friedrich-Alexander-Universität Erlangen-Nürnberg, where my research focuses on computer vision and the foundations of machine learning. I have recently conducted a research project on the “Comparative Analysis of OOD Detection Methods on the Historical-WI Dataset” and am currently working on handwriting imitation in the context of generative modeling.

Alongside my studies, I am employed as a Working Student – Data Quality Manager at CheckMyBus GmbH. I also bring prior industry experience as a working student at BMW Group and Elektrobit Automotive GmbH, complemented by expertise in data analytics, software engineering, and project management. Earlier in my career, I served as Project Coordinator at Robi Axiata Limited and as a Teaching Assistant and Lab Instructor at North South University, where I supported students in developing core computer science skills.

Outside of my professional life, I enjoy volunteering for social welfare causes. In my free time, I love to read literature, including mysteries, fantasies, science fiction, and classics. When I’m not immersed in the digital world, I like to go cycling in the countryside or take peaceful walks along serene roads.

Feel free to reach out if you are interested in collaborating on any research or exchanging ideas.

Profiles: GitHub | LinkedIn | ResearchGate | Twitter | Goodreads

news

Oct 15, 2024	Joined as a Working Student – Data Quality Manager at CheckMyBus GmbH, Nürnberg, DE.
Jul 10, 2023	Joined as a Working Student - Data Analyst at BMW Group, Munich, DE.
Jul 05, 2022	Joined as a Working Student - Software Engineer in Automation at Elektrobit Automotive GmbH, Erlangen, DE.

latest posts

Aug 18, 2025	তন্ময়কে লেখা খোলা চিঠি
Aug 17, 2025	ছেলেটা
May 22, 2021	a distill-style blog post

selected publications

ICBSLP
Bangla Short Speech Commands Recognition Using Convolutional Neural Networks

Shakil Ahmed Sumon, Joydip Chowdhury, Sujit Debnath, and 2 more authors

In 2018 International Conference on Bangla Speech and Language Processing (ICBSLP), 2018

Abs DOI Bib

Despite being one of the most widely spoken languages of the world, no significant efforts have been made in Bangla speech recognition. Speech recognition is a difficult task, particularly if the demand is to do so in noisy real-life conditions. In this study, Bangla short speech commands data set has been reported, where all the samples are taken in the real-life setting. Three different convolutional neural network (CNN) architectures have been designed to recognize those short speech commands. Mel-frequency cepstral coefficients (MFCC) features have been extracted from the audio files in one approach whereas only the raw audio files have been used in another CNN architecture. Lastly, a pre-trained model which is trained on a large English short speech commands data set has been fine-tuned by retraining on Bangla data set. Experimental results reveal that the MFCC model shows better accuracy in recognizing Bangla short speech commands where, surprisingly, the model predicting on raw audio data is very competitive. The models have shown proficiency in identifying single syllable words but encounter difficulties in recognizing multi-syllable commands.
@inproceedings{8554395, author = {Ahmed Sumon, Shakil and Chowdhury, Joydip and Debnath, Sujit and Mohammed, Nabeel and Momen, Sifat}, booktitle = {2018 International Conference on Bangla Speech and Language Processing (ICBSLP)}, title = {Bangla Short Speech Commands Recognition Using Convolutional Neural Networks}, year = {2018}, volume = {}, number = {}, pages = {1-6}, keywords = {Speech recognition;Mel frequency cepstral coefficient;Feature extraction;Data models;Convolutional neural networks;Hidden Markov models;Automatic Speech Recognition;Bangla Speech Recognition;Short Speech Commands;MFCC;Transfer learning;Convolutional neural network}, doi = {10.1109/ICBSLP.2018.8554395}, }
IS
Early Detection of Glaucoma Using Fuzzy Logic in Bangladesh Context

Nazmul Alam Diptu, Md. Asif Khan, Sujit Debnath, and 4 more authors

In 2018 International Conference on Intelligent Systems (IS), 2018

Abs DOI Bib

Detecting Glaucoma at an early stage is very crucial to prevent irreversible vision loss. Glaucoma detection often requires advanced diagnosis which are not accessible to most of the hospitals in a developing country like Bangladesh. Therefore, detection of Glaucoma with the help of common and fewer tests would certainly improve the condition of the Glaucoma patients. In this paper, a method is devised to detect Glaucoma with the data obtained from two ophthalmological tests which are OCT and tonometry along with some other risk factors. The method uses Adaptive Neuro-Fuzzy Inference System (ANFIS) to train an Artificial Intelligence model which can make prediction about the presence of Glaucoma, the absence of it and whether the patient is suspected to have Glaucoma. Conventional Glaucoma detection techniques suggest the use of Intraocular pressure (IOP) as one of the primary parameters to detect Glaucoma. But by analyzing data of Glaucoma patients of Bangladesh, we observed that IOP is not the only significant factor for detecting Glaucoma. Our proposed method gives an accuracy of 81.25% using the ANFIS model.
@inproceedings{8710490, author = {Diptu, Nazmul Alam and Khan, Md. Asif and Debnath, Sujit and Imam, Abdullah Al and Rakib, Al Mahadi Hasan and Ahmed Ador, Kazi Asfaq and Rahman, Rashedur M.}, booktitle = {2018 International Conference on Intelligent Systems (IS)}, title = {Early Detection of Glaucoma Using Fuzzy Logic in Bangladesh Context}, year = {2018}, volume = {}, number = {}, pages = {87-93}, keywords = {Erbium;Magnetic resonance imaging;Hafnium;ANFIS;Early Glaucoma Detection;FIS;Glaucoma;OCT;Vision Loss}, doi = {10.1109/IS.2018.8710490}, }