A collection of my research projects, covering computer vision, NLP, generative AI, and healthcare applications.
Multi-stage pipeline for Braille recognition using YOLOv11, PCA deskewing, and attention-based CNN with transformer correction. Achieved 9.7% CER and eliminated silent decoding failures.
Hybrid evolutionary-adversarial model combining YOLO detection, PSO/CMA-ES refinement, and dual-GAN architecture for realistic handwriting generation.
Transformer-driven pipeline for emotion-aware translation across South Indian languages using XLM-RoBERTa and metric-guided MarianMT decoding.
Advanced Intrusion Detection System for Industrial IoT using LSTM, TCN-Attention, and CNN-LSTM models with flow-based splitting. Achieved 99.99% accuracy.
Comprehensive ML pipeline using neural networks and ensemble methods (XGBoost, LightGBM, CatBoost) for cardiovascular risk assessment across multiple datasets.
DST-SERB funded project for visually challenged users using YOLOv8 hazard detection and fuzzy routing algorithms.