Journals

  • M. Y. Hossain, M. M. H. Rakib, S. Rajit, I. R. Nijhum, R. M. Rahman, Adaptive and automatic aerial image restoration pipeline leveraging pre-trained image restorer with lightweight Fully Convolutional Network, Expert Systems with Applications, 2025 [Paper]

    This work presents a novel pipeline for automatically categorizing and restoring degraded aerial images captured by UAVs, utilizing a lightweight adaptive degradation classifier and a pre-trained transformer-based restoration model. The system achieved high classification accuracy and effective restoration, demonstrating its applicability in vehicle detection and highlighting avenues for future improvements.

Conferences

  • S. Rajit, Z. F. Ananna, M. M. Ehsan, N. N. Punom and S. Siddique, Multi-Class Brain Tumor Classification of MRI Image Using Federated Learning with Blockchain, IEEE Region 10 Symposium (TENSYMP), 2024 [Paper]

    This paper proposes a Federated Learning framework for diagnosing brain tumors while preserving patient data privacy through secure model parameter sharing facilitated by Blockchain technology. Building upon my undergraduate dissertation, this research extends the previous work by using an efficient and accelerated weighted average aggregation method, the framework achieved impressive accuracy in classifying four tumor types, showcasing its efficiency and potential for collaborative medical diagnosis.

  • S. Rajit, M. A. A. Sayed, Federated Learning Based Histopathological Image Classification for Oral Squamous Cell Carcinoma, 8th IEEE-EMBS Conference on Biomedical Engineering and Sciences (IECBES), 2024 [Presented; Proceedings Pending]

    This study introduces a Federated Learning framework for detecting Oral Squamous Cell Carcinoma from histopathological images, enabling collaborative model training while preserving patient data privacy. Utilizing both IID and non-IID datasets along with the Federated Averaging algorithm for weighted parameter distribution, the framework demonstrates its effectiveness in enhancing diagnostic reliability and adapting to diverse data distributions.