Yunfeng Cheng is a graduate student of Intelligent Systems at National University of Singapore. His research interests include machine learning, deep learning, and computer vision.
Interests
- Machine Learning
- Deep Learning
- Computer Vision
Education
MTech in Intelligent Systems, 2021
National Uinversity of Singapore
BEng in Electornic Information Engineering, 2020
Shandong University
Experience
Student
Courses include:
- Intelligent Reasoning Systems
- Pattrern Recognition Systems
- Intelligent Sensing Systems
- Intelligent Robotic Systems
Projects
Tennis Stroke Detection
Using YOWO model to perform tennis strokes detection. Achieve 90% in recall, 89% in classification accuracy, and 81% in localization accuracy.
- First, use 3D-CNN and 2D-CNN to extract spatio-temporal features.
- Second, adopt attention-based mechanism to fuse the extracted features.
- Third, send fused features to mlp to get the bounding box information on the current frame.
Key: 3D-CNN, Attention-based channel fusion
FashionAI - Digitalizing Fashion Retailing
Design 3 different intelligent systems
- Generating fashion articles for design inspirations using GAN.
- Swapping model clothing style for product marketing using CycleGAN.
- Virtual try-on for consumers using Conditional Analogy GAN.
Key: GAN, CycleGAN, Conditional Analogy GAN
No-reference image quality assessment(NR-IQA)
Get MOS/DMOS value of image without reference.
- First, use Siamese Network to rank images according to image quality.
- Second, finetune on traditional IQA dataset to get absulte image quality value.
Key: Siamese network, Learning to rank