Modular PyTorch framework: Pydantic schemas + Optuna optimization + resolution-aware architectures for vision research

Deep learning fish classifier combining ConvNeXt-Tiny (40 species, 98.96% accuracy) with BioCLIP-2 zero-shot recognition and AI-powered habitat mapping

Project using PyTorch in which we create custom datasets and dataloaders, train a convnext_tiny model and log it using tensorboard, do inferences and use Captum for more detailed results.

Mammographic images classification.

This repository contains my work on Alzheimer's Disease detection using deep learning models applied to neuroimaging data. The projects explore multiple architectures and datasets to classify Alzheimer's stages based on MRI scans.

Exploring the Application of Attention Mechanisms in Conjunction with Baseline Models on the COVID-19-CT Dataset

I built a web app for medical image analysis that allows users to upload images and receive classification results. The backend uses Spring Boot with PostgreSQL for authentication and role management, while FastAPI handles the image processing and making predictions. On the frontend, I developed a responsive ReactJS interface.

About Deep learning for ASD / Autism detection

A multi-modal deep learning framework for skin lesion classification on the HAM10000 dataset, combining dermoscopic images and clinical metadata using a ConvNeXt-Tiny backbone to achieve robust performance under class imbalance.

Classifying Brain tumor images using Late fusion of two pre trained cnn model ConvNextTiny , you can test MRI image to classify

YOLO-style object detector implemented from scratch with a custom loss function, pre-trained feature extractor, and end-to-end training pipeline on PASCAL VOC for real-time object localization and classification. Tech: Python (pytorch, scikit-learn, numpy, matplotlib, tqdm, PIL)

Deep learning for ASD / Autism detection