Analyzed laser data for detecting failure mode in the LCD production process by performing autoencoders in PyTorch. Classi ed the good and defective products, accelerated defect detection, and reduced 31% scrap rate. Initiated a project applying real-time object detection to judge the correctness of components on the robot arm. Implemented data augmentation with OpenCV after collecting images using Raspberry Pi. Trained the model with YOLOv3 and detected objects with 96% accuracy to ensure the correctness of installation.

Developed an automatic production line to advance industrial traceability with a team of 4. Integrated computer vision and a robotic arm to realize the randomly placed workpiece picking up. Optimized the laser marking controller for recording information of workpieces by modifying the PLC ladder diagram.