Shaopeng Guo

I am a research assitant at HKUST supervised by Prof. Chi-Keung Tang and Prof. Yu-Wing Tai , where I work on computer vision and machine learning. I also work remotely with Prof. Cewu Lu.

I also obtained Bachelor’s degree at HKUST in 2018. From 2018 to July 2021, I worked as a full-time R&D in a leading Chinese Tech. Company for three years.

Email  /  Google Scholar  /  CV  /  Github

Instead of directly applying discrete human-part labels during classification, we incorporate a language model BERT to transform natural language labels to continuous latent feature vectors and train our model by contrastive loss.

We propose a framework to train an efficient visual Transformer. With the help of the proposed patch compression mechanism, the model can inference with about 2x faster speed and equal or better performance.

We propose a new Convolution-enhanced image Transformer (CeiT) which combines the advantages of CNNs in extracting lowlevel features, strengthening locality, and the advantages of Transformers in establishing long-range dependencies.

We propose a method called Differentiable Dynamic Wirings (DDW), which learns the instance-aware connectivity that creates different wiring patterns for different input images.

We model the channel pruning as a Markov process, in which each state represents for retaining the corresponding channel during pruning.

The template of this website is from Jon Barron.