ProvRain introduces a rain-adaptive detection pipeline using a lightweight MobileNet-U-Net denoiser and Faster R-CNN classifier. Designed for nighttime driving under severe rain, the method improves recall, accuracy, and early-warning success while remaining efficient enough for real-time automotive hardware.

PROXI-NAV is a proximity-aware navigation system combining a multi-head YOLOv8 network for obstacle detection and pavement segmentation with FastDepth-based monocular depth estimation. Deployed on the NVIDIA Jetson Nano, it delivers low-latency real-time auditory navigation feedback for the visually impaired, enabling safer and more autonomous mobility.

CLAIMS is an explainable clinical ECG interpretation framework combining rule-based signal features, CNN embeddings, and LLM-guided narrative synthesis. Designed for the PTB-XL dataset, it generates transparent diagnostic reports that include evidence citations, improving trustworthiness and interpretability over traditional black-box deep learning models.