Ongoing
Postdoctoral research on drunk/impaired driver detection using thermal (LWIR) cameras, VLM-based evaluation pipelines, and open-vocabulary detection for driver monitoring.
Completed
27-camera smart factory human intrusion and defect detection system, road anomaly detection, gaze estimation, and model deployment on Jetson, TI, and Hailo hardware.
Multi-camera 3D perception, monocular depth-based pseudo-LiDAR.
Real-time deep learning systems for driver and occupancy monitoring, seatbelt detection, drowsiness, and ADAS features deployed on edge hardware.
GAN-based research on few-shot font generation, font family synthesis, and text image skeletonization — with publications in CVPR, Expert Systems with Applications, Knowledge-Based Systems, and IJDAR.
Unsupervised GAN architecture for content-style disentanglement enabling image style transfer, domain translation, and attribute manipulation without label supervision.