Research Projects

• Hyperparameter Analysis For Image Captioning Based On Different Languages (2020):
  • Technology: Python, PyTorch, TensorFlow & Numpy.
  • We analyzed different hyperparameter combinations for two state-of-the-art image captioning models based on different languages: Chinese, Bangla and English and tried to find suitable combinations of hyperparameters with this analysis. This experiment was carried out under the supervision of Dr. Md Shahriar Karim, Assistant Professor, North South University, Bangladesh. [project report, presentation video]
• An Efficient Learning Rate System in Deep Learning (2020):
  • Technology: Python, PyTorch & Numpy.
  • We explored different Learning Rate scheduler fucntions and proposed five different learning rate scheduling to train any deep learning model. This project aims to help in choosing an optimum learning rate, which will acheive higher accuracy and reduce the time of convergence at an optimum point. This project was done under the supervision of Dr. Md Shahriar Karim, Assistant Professor, North South University, Bangladesh.

Academic Projects (2017-2020)

• 2D Sketch Colorization Using GAN:
  • Technology: Python, TensorFlow & Numpy.
  • Sketch colorization can be applied in many scenarios. Our project will mainly focus on the colorization of black and white comics and manga. This will make the content more immersive for the readers.  It can also help artists get a feel for the colored version of their sketches. Code
• Dementia Prediction Using Different Machine Learning Model:
  • Technology: Python, Scikit-learn & Numpy.
  • We used the “MRI and Alzheimers” dataset provided by the Open Access Series of Imaging Studies (OASIS) project. We used this dataset to train SVM, Logistic Regression, Decision Tree, and Random Forest, models. We used accuracy, recall, AUC, and confusion matrix as evaluation metrics. To improve our result, we fine-tuned all our models using the Grid Search method. Code
• Textbin - A text sharing Webapp:
  • Technology: Python, Django, Bootstrap4
  • A text sharing app similar to pastebin Code
• NSU-Social:
  • Technology: Python, Django, Bootstrap4, SQLite
  • A social media app with customizable profile. User can make posts similar to twitter. Users can also create or join groups. Can only be accessed by NSU students. Code
• Healix:
  • Technology: Python, Django, JavaScript, Bootstrap4, SQLite
  • An online platform for doctors and patients. Has Health Blog, E Prescription, Health Profile, Hospital List. Patients can store prescriptions and test reports as PDF. Patients can get e-prescriptions from doctors. Code
• Hospital Management System:
  • Technology: Java, JDBC, Apache Derby, JavaFX
  • A desktop application used to store patients and doctors information of a particular hospital. Patients can also set appointments with doctors. Code
• 16 Bits MIPS CPU:
  • Technology: Python, Logisim
  • Designed a 16bit ISA and progammed an Assembler in Python. CPU simulation was done using Logisim. Code