● Thesis: SensingGAN: A Lightweight GAN for Real-time Rain Removal with Self-attention (IEEE TNNLS'22 Under Review): SensingGAN can effectively sense objects and rain like humans, and restore the details of objects to satisfy the high safety and efficiency requirements of autonomous vehicles. It increases Object Detection accuracy by 3% in rain-in-driving. In comparison with classical de-raindrop GAN, FPS is improved by 13 times (10 ms). - Skills: GAN, Python, Pytorch, Linux - Github: https://github.com/chenabei0421/SensingGAN

● International ICT Innovative Sevices Awards - 2nd place - Automatic Composition based on Emotion of Image (倒聽圖說): Judge the mood of the image based on Color Psychology, and automatically compose music according to the HSV of the viewpoint. Responsible for the automatic composition of music and demonstration of the project. - Skills: C++, MIDI