About Me
Hello, I’m Golam Jilani, a computer science graduate student. I received my bachelor’s degree from Shahjalal University of Science & Technology (SUST) in 2023, majoring in Computer Science & Engineering. My research interests are in the areas of Machine Learning, Computer Vision, Natural Language Processing, and Data Mining. I am interested in deriving meaningful information from texts/images/videos to solve real-world problems.
Education
Shahjalal University of Science & Technology (SUST)
Sylhet, Bangladesh
B.S. in Computer Science & Engineering
May 2023
GPA: 3.51/4.00 (3.73 in the last two years)
Research Experience
Reasearch Assistant
Jul 2022 - Jun 2023
Supervisor: Dr. Sadia Sultana
In the final year of my undergrad studies, I had the privilege of working part-time as a research assistant on a university research project titled "SUFEDB: A facial expression database for emotion recognition". I collected consent from participants and annotated expression images via OpenCV. I had also done preprocessing of images to ensure data uniformity of the dataset. In addition, I had evaluated and analyzed the performance of current state-of-the-art CNN models on the dataset.
Undergraduate Thesis
Jan 2022 - Mar 2023
Development of an Ensemble Learning system for Facial Expression Recognition using smaller CNN models with Transfer Learning
Supervisor: Dr. Sadia Sultana
- Developed and trained an efficient Facial Expression Recognition ensemble learning system using smaller CNN models with transfer learning, which achieved 97.55% accuracy on the benchmark dataset KDEF.
- Used transfer learning and advanced data augmentation to deal with overfitting problems and assessed the performance of Mixup and CutMix data augmentation on our benchmark datasets.
- Validated the effectiveness of our ensemble model with other existing state-of-the-art methods.
Collaborative Research
Ongoing
In recent years, there has been an increasing interest in expressing thoughts and feelings on social media rather than in face-to-face conversation. Social networks, especially Reddit, have emerged as powerful platforms for sharing depression, which can be utilized to analyze the trends of depression. For this, we collected a bulk amount of depression-related data (1M posts) crawling from Reddit and utilized NLTK for text processing and custom functions for text cleaning. We preprocessed the data in a time-based approach: Incremental Window and Sliding Window. We used pre-trained word embeddings like Skip-gram and GloVe for analyzing trends related to depression, leveraging the semantic relationships encoded in word vectors.
Publications
SUFEDB: A facial expression database for emotion recognition
Sadia Sultana, Saiful Sagor, Golam Jilani, Al Masum, Samara Paul
(under review)
Sadia Sultana, Saiful Sagor, Golam Jilani, Al Masum, Samara Paul
(under review)
An efficient ensemble learning model integrating multi-branch sub-networks for facial expression recognition.
Golam Jilani, Samara Paul, Sadia Sultana
(under review)
Golam Jilani, Samara Paul, Sadia Sultana
(under review)
Technical Skills
- Languages & Databases: Python, C, Java, JavaScript, MySQL, MongoDB.
- Frameworks: MERN Stack (MongoDB, Express.js, React.js, Node.js), Django, LaTeX.
- ML Frameworks: PyTorch, Keras, Tensorflow, Numpy, Pandas, scikit-learn, OpenCV, NLTK, Gensim.