Yang Liu MRUIL

`>_ whoami`

class YangLiu(Researcher):
    education   = ["PhD, King's College London — School of Biomedical Engineering & Imaging Sciences (2021–2026)",
                   "MSc, Huazhong University of Science and Technology (2019–2021), advised by Prof. Xiang Bai"]
    experience  = ["Proximie — Data Scientist, part-time (Jan–Jul 2025)",
                   "ByteDance — Computer Vision Algorithm Engineer, intern (2021)"]
    advisors    = {"primary": "Prof. Sébastien Ourselin", "co-supervisors": ["Prof. Prokar Dasgupta", "Dr. Alejandro Granados"]}

    research    = {
        "surgical video":   ["online phase recognition", "long-video transformers", "streaming VLMs"],
        "segmentation":     ["weakly-supervised referring segmentation", "unsupervised instrument segmentation"],
        "medical imaging":  ["echocardiography phase detection", "stroke lesion segmentation", "AD diagnosis from sMRI"],
        "vision-language":  ["language-driven visual tasks", "noise-injected cross-modal alignment"],
    }

    def mission(self):
        return "Build AI that watches surgery in real time and makes the operating room safer."

`>_ research highlights`

🩺 Surgical Video Understanding

Online, causal recognition of what is happening in the OR — from a single frame to a multi-hour procedure.

StableSPR · MICCAI 2026 (early accept, top 9%) — temporal error-cascade loss + evidence-gated transition predictor for stable online phase recognition
LoViT · Medical Image Analysis 2024 — long video transformer for surgical phase recognition
SKiT · ICCV 2023 — fast key-information video transformer, real-time online inference
Motion-boundary unsupervised instrument segmentation · MICCAI 2025

🧠 Vision-Language & Segmentation

Learning with fewer labels, across modalities.

WeakMCN · CVPR 2025 — multi-task collaborative network for weakly-supervised referring expression comprehension & segmentation
ArcSin · language-driven visual tasks via adaptive-ranged cosine similarity noise injection
Super-BPD · CVPR 2020 — boundary-to-pixel direction for fast image segmentation (~25 fps)
WDNet · WACV 2021 — watermark-decomposition network + the CLWD dataset

🏆 1st place — MICCAI 2022 ATLAS Ischemic Stroke Lesion Segmentation Challenge (Team CTRL)

`>_ publications`

Year	Paper	Links
2026	Stabilizing Temporal Inference Dynamics for Online Surgical Phase Recognition — Yang Liu, Ning Zhu, J. Peng, X. Chen, A. Granados, G. Wang, S. Ourselin	arXiv · code
2026	SLIM: High-Throughput Watermarking with Stationary Latent Manifolds — Q. Yan, Z. Chen, Yang Liu†, Z. Cai†	—
2025	Motion-Boundary-Driven Unsupervised Surgical Instrument Segmentation in Low-Quality Optical Flow — Yang Liu, P. Wu, J. Huo, G. Zhang, Z. Yuan, C. Bergeles, R. Sparks, P. Dasgupta, A. Granados, S. Ourselin	arXiv
2025	WeakMCN: Multi-task Collaborative Network for Weakly Supervised Referring Expression Comprehension and Segmentation — Yang Liu, Silin Cheng, X. He, S. Ourselin, L. Tan, G. Luo	arXiv · code
2025	A Domain Generalization Framework Based on Wavelet-Driven Structural Enhancement and Contrastive Alignment — Y. Xu, T. Zhang, Yang Liu	—
2024	LoViT: Long Video Transformer for Surgical Phase Recognition — Yang Liu, M. Boels, L. C. Garcia-Peraza-Herrera, T. Vercauteren, P. Dasgupta, A. Granados, S. Ourselin	paper · code
2024	Gray Matter-Guided Attention Network for AD Diagnosis Using Structural MRI — Y. Zhang, H. Cai, Y. Du, B. Xu, Yang Liu†	paper
2024	DDSB: An Unsupervised and Training-free Method for Phase Detection in Echocardiography — Z. Bu, Yang Liu†, et al.	arXiv · code
2024	ArcSin: Adaptive ranged cosine Similarity injected noise for Language-Driven Visual Tasks — Yang Liu, X. Yu, G. Zhang, Z. Zhu, C. Bergeles, P. Dasgupta, A. Granados, S. Ourselin	arXiv
2023	SKiT: a Fast Key Information Video Transformer for Online Surgical Phase Recognition — Yang Liu, J. Huo, J. Peng, R. Sparks, P. Dasgupta, A. Granados, S. Ourselin	paper · repo
2021	WDNet: Watermark-Decomposition Network for Visible Watermark Removal — Yang Liu, Z. Zhu, X. Bai	paper · code
2020	Super-BPD: Super Boundary-to-Pixel Direction for Fast Image Segmentation — J. Wan, Yang Liu, D. Wei, X. Bai, Y. Xu	paper · code

_{* equal contribution · † corresponding author}

`>_ featured repositories`

`>_ tech stack`

Deep Learning

Vision · Medical · Multimodal

Infra · Tooling

`>_ github analytics`

`>_ timeline`

2026.05  ◆ MICCAI 2026 early accept (top 9%) — StableSPR          ◆ IEEE TCSVT — SLIM
2026.01  ◆ PhD awarded, King's College London
2025     ◆ CVPR 2025 — WeakMCN   ◆ MICCAI 2025   ◆ ICME 2025   ◆ Data Scientist @ Proximie (Jan–Jul)
2024     ◆ Medical Image Analysis — LoViT   ◆ ISBI 2024   ◆ DDSB / ArcSin preprints
2023     ◆ ICCV 2023 — SKiT
2022     ◆ 1st place, MICCAI ATLAS Stroke Lesion Segmentation Challenge
2021     ◆ MSc, HUST   ◆ CV Algorithm Engineer intern @ ByteDance   ◆ PhD begins @ KCL, London  
2020–21  ◆ CVPR 2020 — Super-BPD   ◆ WACV 2021 — WDNet

Open to collaboration on surgical AI, medical video understanding and vision-language models.
_{yang.9.liu@kcl.ac.uk}

Navigation Menu