CSE undergraduate with research interests in Deep Learning, Computer Vision, Explainable AI, and Applied Machine Learning.
My work focuses on building practical AI systems for real-world problems, particularly in agricultural computer vision, medical imaging, object detection, and interpretable deep learning. I am interested in developing models that are not only accurate, but also efficient, explainable, and reliable across different environments.
Deep Learning • Computer Vision • Explainable AI • Object Detection •
Agricultural AI • Medical Image Analysis • Domain Generalization • Representation Learning
Explainable Ensemble Deep Learning for Sugarcane Leaf Disease Classification
An explainable deep-learning framework combining pretrained CNNs, SE/CBAM attention mechanisms, weighted soft voting, and Grad-CAM for 14-class sugarcane leaf disease recognition.
- Final ensemble accuracy: 97.20%
- F1-score: 97.22%
- Accepted and presented at IEEE RAAICON 2026
Cross-Domain Wheat Head Detection and Counting on GWHD 2021
A comparative study of YOLOv8 and YOLO26 for wheat-head detection, cross-domain generalization, computational efficiency, and image-level counting.
- F1-score: 0.763
- mAP@0.50: 0.762
- Counting MAE: 8.085
- Accepted for presentation at IEEE WIECON-ECE 2026
Interpretable Lumbar Spinal MRI Classification via Self-Supervised Learning
A medical imaging study exploring supervised, self-supervised, and semi-supervised learning using BYOL, DINO, and STAC, together with Grad-CAM and occlusion sensitivity for interpretability.
- BYOL-VGG16 accuracy: 95.63%
- F1-score: 95.62%
- Accepted for oral presentation as a regular paper at IEEE CSNT 2027
Languages: Python, C/C++, Java
Deep Learning: TensorFlow, Keras, PyTorch
Computer Vision: OpenCV, Ultralytics YOLO
Machine Learning: Scikit-learn
Research Tools: Jupyter Notebook, Google Colab
Development: Git, GitHub, Streamlit
I am currently strengthening my research experience in computer vision and trustworthy AI, with particular interest in explainable learning systems, robust visual recognition, and research-driven applications of deep learning.
- GitHub: mahfuz1715
- LinkedIn: Mahfuz Uddin Ahmed