Shengju Qian

I am a PhD student at The Chinese University of Hong Kong, supervised by Prof. Jiaya Jia.

Prior to that, I received my Bachelor degree from the University of Electronic Science and Technology of China. I've spent wonderful time with my internship at Amazon AI, Tencent Youtu X-Lab, and SenseTime Research.

My research interests lie primarily in computer vision and deep learning, particularly focusing on representaion learning and network designs, including self-supervised learning, transformers for vision, and generative models for face/body understanding.

Email / LinkedIn / Google Scholar / Github

Blending Anti-Aliasing into Vision Transformer
Shengju Qian, Hao Shao, Yi Zhu, Mu Li, Jiaya Jia
Neural Information Processing Systems (NeurIPS) , 2021
[pdf] [code] [bibtex]

We explore anti-aliasing strategies for vision transformers to mitigate the drawback of 'patch-wise' tokenization.

Temporal Interlacing Network
Hao Shao, Shengju Qian, Yu Liu
AAAI Conference on Artificial Intelligence (AAAI), 2020
[pdf] [code] [bibtex]

We propose an efficient and accurate framework for video recognition with TIN, which has a strong capability of temporal modeling.

Aggregation via Separation: Boosting Facial Landmark Detector with Semi-Supervised Style Translation
Shengju Qian, Keqiang Sun, Wayne Wu, Chen Qian, Jiaya Jia
IEEE International Conference on Computer Vision (ICCV), 2019
[pdf] [code] [bibtex]

We propose a semi-supervised learning algorithm for boosting facial landmark detectors.

Make a Face: Towards Arbitrary High Fidelity Face Manipulation
Shengju Qian, Kwan-Yee Lin, Wayne Wu, Chen Qian, Ran He et al
IEEE International Conference on Computer Vision (ICCV), 2019
[pdf] [project page] [bibtex]

We propose Addtive Focal Variational Autoencoder(AF-VAE) for high fidelity face manipulation, which is robust to arbitrary shape variation.

Extending the Capacity of CVAE for Face Synthesis and Modeling
Shengju Qian, Wayne Wu, Yangxiaokang Liu, Beier Zhu, Fumin Shen
NeurIPS Workshop on Relational Representation Learning Workshop (NIPSW), 2018
[pdf] [bibtex]

Presenting a practical methodology that can boost CVAE's representation power on face synthesis and modeling.