|
I am a PhD student at the Visual Geometry Group (VGG) at University of Oxford, advised by Andrew Zisserman and Weidi Xie, where I work on computer vision, more specifically video understanding. I am generously funded by EPSRC and AIMS CDT. I also spent half a year interning at Meta Reality Labs. I did my undergraduate in Engineering Science, also at Oxford. During which, I spent lovely summers at Japan Railways, Metaswitch (now acquired by Microsoft), True, CP Group, and Oxford’s Engineering Department. Prior to which, I was born and raised in the suburbs of Bangkok, Thailand. Email / CV / Twitter / Github / Google Scholar |
|
|
My research focuses on computer vision. I am particularly interested in: video understanding, self-supervised learning, segmentation, and new applications. |
|
Charig Yang PhD Thesis abstract / thesis (coming soon) Submitted my thesis! This thesis explores novel methods in learning and using temporal signals in videos. I will be defending in April. My examiners will be Christian Rupprecht (Oxford) and Bill Freeman (MIT). |
|
Charig Yang, Weidi Xie, Andrew Zisserman ECCV, 2024 (Oral Presentation) project page / arXiv Changes happen all the time, but only some are consistent over time. Ordering shuffled sequences reveals the latter. |
|
Junyu Xie, Charig Yang, Weidi Xie, Andrew Zisserman ACCV, 2024 (Oral Presentation) project page / arXiv SAM + Optical Flow = FlowSAM. |
|
Charig Yang, Weidi Xie, Andrew Zisserman CVPR, 2022 project page / arXiv / tweet (by Lucas Beyer) / new scientist article We solve a niche but fun problem of reading clocks (that 2025's VLMs still fails!). We circumvent manual supervision by exploiting the fact that time flows at a constant rate. |
|
Charig Yang, Hala Lamdouar, Erika Lu, Andrew Zisserman, Weidi Xie Short: CVPR Workshop on Robust Video Scene Understanding, 2021 (Best Paper Award) Full: ICCV, 2021 project page / arXiv Motion can be used to discover moving objects in general. We introduce a self-supervised segmentation method by grouping motion into layers using a transformer. |
|
Hala Lamdouar, Charig Yang, Weidi Xie, Andrew Zisserman ACCV, 2020 project page / arXiv Camouflaged animals are hard to see, only until they move. We present a method of discovering camouflages using motion, and a large-scale video camouflage dataset. |
|
2022-23: A2 (second-year) Electronic and Information Engineering, B14 (third-year) Information Engineering Systems, C18 (fourth-year) Computer Vision and Robotics
You can find my summary notes for all P and A modules, and some B modules here. |
|
Template gratefully stolen from here. |