Ching-Jui Wang

My name is Ching-Jui Wang, and I'm currently a student at the graduate school of Electrical Engineering from National Taiwan University. The major research of mine is computer vision, including human body fields like 2D/3D human pose estimation. Based on my major research, I'm pretty familiar with ML/DL knowledge and techniques like python, pytorch etc. Moreover, I'm also fascinated with the application and knowledge of firmware/software systems, which encourages me to study C/C++ and applications on OS and Algorithm. Last but not least, in the generation of AI, I'm also familiar with using the applications of AI tools, not only for improvements on my personal projects, but also improve the efficiency.

• Personal Research: spatial-temporal GCN-Transformer model for 3D Human Pose Estimation

Integrates a graph convolutional network (GCN) and attention mechanisms, operating across both spatial and temporal domains. Starting with 2D poses extracted from RGB images, the approach involves lifting these stored 2D pose sequences into 3D, utilizing a motion-aware encoder and decoder that refine the poses by incorporating positional, velocity, and acceleration features.

• Industry-academic collaboration project (Jorjin Ltd.): Human reconstruction for AR/VR applications.

In order to fulfill the goal for more vivid online meetings or teaching scenarios, we extract human body skeleton features and appearances for human reconstruction in the virtual reality world as an independent avatar. I majored in the applications of detecting human body parts and gestures, also integrating them with spatial and temporal features for better reconstruction in Unity.