Shuyang (Kevin) Sun

I am currently a research scientist at ByteDance. Previously I got my DPhil (PhD) degree at University of Oxford under the supervision of Professor Philip Torr and Professor Victor Prisacariu. I received my M.Phil. degree from the University of Sydney, where I was with the SIGMA Lab, School of Electrical & Information Engineering, supervised by Professor Wanli Ouyang. I am fortunate to collaborate closely with Weijun Wang, and Liang-Chieh Chen at Google Research, Vladlen Koltun, Philipp Krähenbühl and René Ranftl at Intel ISL. I received a B.Eng. degree in 2016 in Software Engineering from Wuhan University, China.

Personal email: dr.shuyang.sun@gmail.com

news

Jul 11, 2022	Internship started at Google Research.

Selected Publications

For the full publication list, please refer to my Google Scholar.

         
CLIP as RNN: Segment Countless Visual Concepts without Training Endeavor
 Shuyang Sun*, Runjia Li*, Philip Torr, Xiuye Gu, and Siyang Li
 In CVPR, 2024 
 
 Bib PDF Code Website 
 @inproceedings{clip_as_rnn,
  title = {CLIP as RNN: Segment Countless Visual Concepts without Training Endeavor},
  author = {Sun*, Shuyang and Li*, Runjia and Torr, Philip and Gu, Xiuye and Li, Siyang},
  year = {2024},
  booktitle = {CVPR},
  selected = true,
  pdf = https://arxiv.org/pdf/2312.07661.pdf,
  website = https://torrvision.com/clip_as_rnn/,
  code = https://github.com/kevin-ssy/clip_as_rnn,
  preview = clip_as_rnn.jpg,
  highlight = true
} 

ReMaX: Relaxing for better training on efficient panoptic segmentation

Shuyang Sun, Weijun Wang, Qihang Yu, Andrew Howard, Philip Torr, and Liang-Chieh Chen

NeurIPS, 2023

Bib PDF Website

@article{sun2023remax,
  title = {ReMaX: Relaxing for better training on efficient panoptic segmentation},
  author = {Sun, Shuyang and Wang, Weijun and Yu, Qihang and Howard, Andrew and Torr, Philip and Chen, Liang-Chieh},
  journal = {NeurIPS},
  year = {2023},
  selected = true,
  pdf = https://arxiv.org/pdf/2306.17319.pdf,
  website = https://nips.cc/virtual/2023/poster/71598,
  preview = remax.jpg
}

Real-Fake: Effective Training Data Synthesis Through Distribution Matching

Jianhao Yuan, Jie Zhang, Shuyang Sun, Philip Torr, and Bo Zhao

ICLR, 2024

Bib PDF Code Website

@article{yuan2023real,
  title = {Real-Fake: Effective Training Data Synthesis Through Distribution Matching},
  author = {Yuan, Jianhao and Zhang, Jie and Sun, Shuyang and Torr, Philip and Zhao, Bo},
  journal = {ICLR},
  year = {2024},
  selected = true,
  preview = realfake.jpg,
  website = https://torrvision.com/realfake/,
  pdf = https://arxiv.org/abs/2310.10402,
  code = https://github.com/baai-dcai/training-data-synthesis
}

Is synthetic data from generative models ready for image recognition?

Ruifei He, Shuyang Sun, Xin Yu, Chuhui Xue, Wenqing Zhang, Philip Torr, Song Bai, and Xiaojuan Qi

ICLR, spotlight, 2023

Bib PDF Code

@article{he2022synthetic,
  title = {Is synthetic data from generative models ready for image recognition?},
  author = {He, Ruifei and Sun, Shuyang and Yu, Xin and Xue, Chuhui and Zhang, Wenqing and Torr, Philip and Bai, Song and Qi, Xiaojuan},
  journal = {ICLR, spotlight},
  year = {2023},
  selected = true,
  code = https://github.com/cvmi-lab/syntheticdata,
  pdf = https://arxiv.org/abs/2210.07574,
  preview = synthesis.png
}

TransMix: Attend to Mix for Vision Transformers

Shuyang Sun*, Jie-Neng Chen*, Ju He, Philip Torr, Alan Yuille, and Song Bai

CVPR, 2021

Bib PDF Code

@article{chen2021transmix,
  title = {TransMix: Attend to Mix for Vision Transformers},
  author = {Sun*, Shuyang and Chen*, Jie-Neng and He, Ju and Torr, Philip and Yuille, Alan and Bai, Song},
  journal = {CVPR},
  year = {2021},
  selected = true,
  pdf = https://arxiv.org/pdf/2111.09833.pdf,
  code = https://github.com/beckschen/transmix,
  preview = transmix.jpg
}

Visual Parser: Representing Part-whole Hierarchies with Transformers

Shuyang Sun, Xiaoyu Yue, Song Bai, and Philip Torr

arXiv preprint arXiv:2107.05790, 2021

Bib PDF Code

@article{sun2021visual,
  title = {Visual Parser: Representing Part-whole Hierarchies with Transformers},
  author = {Sun, Shuyang and Yue, Xiaoyu and Bai, Song and Torr, Philip},
  journal = {arXiv preprint arXiv:2107.05790},
  year = {2021},
  selected = true,
  preview = vip.png,
  pdf = https://arxiv.org/abs/2107.05790,
  code = https://github.com/kevin-ssy/vip
}

Vision transformer with progressive sampling

Xiaoyu Yue*, Shuyang Sun*, Zhanghui Kuang, Meng Wei, Philip Torr, Wayne Zhang, and Dahua Lin

In ICCV, 2021

Bib PDF Code

@inproceedings{yue2021vision,
  title = {Vision transformer with progressive sampling},
  author = {Yue*, Xiaoyu and Sun*, Shuyang and Kuang, Zhanghui and Wei, Meng and Torr, Philip and Zhang, Wayne and Lin, Dahua},
  booktitle = {ICCV},
  pages = {387--396},
  year = {2021},
  selected = true,
  preview = psvit.png,
  pdf = https://arxiv.org/abs/2108.01684,
  code = https://github.com/yuexy/ps-vit
}

Fishnet: A versatile backbone for image, region, and pixel level prediction

Shuyang Sun, Jiangmiao Pang, Jianping Shi, Shuai Yi, and Wanli Ouyang

NeurIPS, 2018

Bib PDF Code

@article{sun2018fishnet,
  title = {Fishnet: A versatile backbone for image, region, and pixel level prediction},
  author = {Sun, Shuyang and Pang, Jiangmiao and Shi, Jianping and Yi, Shuai and Ouyang, Wanli},
  journal = {NeurIPS},
  volume = {31},
  year = {2018},
  selected = true,
  pdf = https://arxiv.org/pdf/1901.03495.pdf,
  code = https://github.com/kevin-ssy/fishnet,
  preview = fish.png
}

Optical Flow Guided Feature: A Fast and Robust Motion Representation for Video Action Recognition

Shuyang Sun, Zhanghui Kuang, Lu Sheng, Wanli Ouyang, and Wei Zhang

In CVPR, 2018

Bib PDF Code

@inproceedings{sun2018optical,
  title = {Optical Flow Guided Feature: A Fast and Robust Motion Representation for Video Action Recognition},
  author = {Sun, Shuyang and Kuang, Zhanghui and Sheng, Lu and Ouyang, Wanli and Zhang, Wei},
  booktitle = {CVPR},
  year = {2018},
  selected = true,
  pdf = https://arxiv.org/pdf/1711.11152.pdf,
  code = https://github.com/kevin-ssy/optical-flow-guided-feature,
  preview = off.jpg
}