Projects
Hybrid LLM Navigation System
Context-aware ROS2 navigation that selects driving policies in crowded environments — 82% vs 43% goal-arrival success over the Nav2 baseline.
MobileViT Motorcycle Traffic-Violation Detection
Temporal traffic-event recognition with YOLO and MobileViT, deployed in real time on a Jetson Orin NX.
ROS2 UGV Platform Integration
Autonomous-driving platform integration for an unmanned ground vehicle: sensors, CAN communication, SLAM, and navigation.
LLM-Assisted Strawberry Quality Assessment
Smart-farm system linking segmentation and quantitative image analysis to RAG and LLM-generated quality reports.
Few-Shot Writer Identification with Vision Transformers
ViT-based prototypical meta-learning that identifies writers from a few handwriting samples — extended into a patent application.
Deep-Learning Camera Fault Detection
Deep-learning pipeline that detects and cleans camera noise and environmental artifacts during drone operation.