Members

This study develops the iEar intelligent eardrum lesion recognition system, combining object detection and image segmentation to locate lesions, analyze contours, and support real-time clinical decisions.

This study uses low-resolution thermal images and human skeleton recognition to detect head-related self-harm behaviors in real time, balancing privacy protection, accurate alerts, and psychiatric-care safety.

This study combines ToF sensing and deep learning to recognize nursing actions. YOLO and hand-skeleton recognition identify turning and back-patting actions to support care quality and patient safety.

This study builds the reTon tongue-muscle rehabilitation game platform, using YOLO object detection and tracking to convert tongue movements into interactive game control for rehabilitation training.

This study combines YOLO object detection with custom algorithms to build an intelligent surgical instrument recognition and training platform for standardized medical education and objective learning assessment.

This study proposes PoseAgent, a VLM agent framework that automatically annotates pig skeletal keypoints through multiple annotator agents and an expert-agent review mechanism, establishing pig-specific OKS tolerances and a skeleton dataset to reduce manual annotation cost and support subsequent behavior monitoring applications.

This study develops a dual-station AI inspection system based on deep learning object detection, which verifies the type and quantity of seasoning packets and detects hair foreign objects on noodle blocks. By integrating an edge computing platform with PLC control, the system performs real-time determination and triggers alerts while the production line is running, establishing an automated quality inspection workflow deployable in food packaging facilities.