Shing Hin Hung

Trained CNN models such as YOLOv8 to detect crops and distinguish crops from weeds. Explored deployment on AI accelerators (e.g., Hailo-8) and packaging solutions with VMs and Docker for portability. Investigated CNN and Vision Transformer models (e.g., SCDepth v3 and Depth Anything v2) for human detection and monocular depth estimation in industrial safety applications. Collected data, performed augmentation and filtering, and conducted training, validation, and testing of fine-tuned models. Worked with ML frameworks including TensorFlow, PyTorch, and CUDA to develop and accelerate AI models.

Designed, developed, and configured robotic systems to meet business process and application requirements. Built a high-speed 4-axis robot arm with control software and data communication for automated labeling. Created a Python-based machine-vision system for 2D object recognition, including local feature extraction and matching. Designed a C#-based HMI for customers.

Developed and constructed automated systems, including testing, maintenance, and data collection. Supported vision-sensor integration with built-in AI. Trained vision systems to cluster objects and determine accurate pick points for robot arms. Programmed PLCs and planned paths for 6-axis robot arms to control and collaborate with other devices. Designed and implemented HMIs for automation systems. Built machine-vision programs in Python using OpenCV, including 3D reconstruction and 3D image processing tasks such as smoothing, denoising, and segmentation.