Amrita University
Research Work
2024 - 2025
Technologies Used:
- Developed a lightweight, explainable Conformer-based model that combines CNN and Transformer blocks for accurate medical image classification across datasets like BloodMNIST, RetinalMNIST, and ROP.
- Demonstrated superior performance (AUC: 0.999 on BloodMNIST) and scalable deployment potential in low-resource settings compared to traditional CNNs and Vision Transformers.
