Computer Vision Researcher. Associate Professor at SJTU, Previously @Oxford_VGG. 中文名：谢伟迪 Personal Webpage: https://teapaw.org/t.co/sZoZ0AfKrX
Oxford, England
Joined May 2018
- Tweets577
- Following628
- Followers3K
- Likes1K
Weidi Xie retweeted
What does it take to shape the future of science? It starts with asking genuine, novel and meaningful scientific questions.
At the Young Scientists Forum of 2026 WLF, 80+ young researchers joined leading scientists to explore how science can reshape the next decade — across AI, materials science, life sciences and sustainability.
The forum also spotlighted the inaugural WLF Prize for Young Scientist Laureates — Raphaële Juliette CLÉMENT, CHEN Peining, XIE Weidi (@WeidiXie) and DING Shiqian — who took the stage to share their latest breakthroughs.
In dialogue with leading scientists, they also explored AI's growing role in research — and why human creativity and critical thinking remain essential to discovery.
A new generation is defining the questions that could lead to tomorrow's breakthroughs!
Weidi Xie retweeted
🤖 Can AI help bring new hope to patients with rare diseases?
XIE Weidi (@WeidiXie) of Shanghai Jiao Tong University (@sjtu1896), a laureate of the inaugural WLF Prize for Young Scientists, is working at the intersection of artificial intelligence and healthcare, with research focused on DeepRare.
In his acceptance remarks, Prof. XIE thanked his mentors, collaborators, students and team, and highlighted the potential of AI to help address diseases that remain difficult to treat.
He hopes his team can continue exploring how AI-driven discoveries can bring better prospects to patients.
For Prof. XIE, the promise of AI lies not only in technological progress, but also in its potential to make a difference in people's lives.
Weidi Xie retweeted
The 2026 World Laureates Forum will be held in Shanghai's Lin-gang Special Area from Oct 30 to Nov 1 under the theme "New Science Paradigm: Break Through for the Next Decade". Click the link to see what's on the agenda!👉 brnw.ch/21x6iWJ @wlf_fdnofficial @QingyaoAi @WeidiXie
![[Photo/VCG]](https://pbs.twimg.com/media/HUQZSs_boAAsTL0.jpg?name=orig)
ALT [Photo/VCG]
Excited to present OmniStream at ECCV —a unified causal backbone that perceives, reconstructs, and acts from continuous video. 🧐🧐
One frozen model transfers across 2D/3D perception, video-spatial reasoning, and robot manipulation.
go2heart.github.io/omnistrea…
11
1,115
Applying AI for sports understanding !
🏀 Can AI truly understand who did what, and when in a basketball game?
Introducing BasketEvent+PlayNet: player-centric basketball understanding of real NBA broadcasts, a new work with @WeidiXie.
🎥 Demo attached. Paper, code&data are now open-source.
🌐 zhangyu2003.github.io/Basket…
SLoMO !
🎬 SLoMO is back!
Ready to push the frontier of story-level movie understanding and Audio Description?
Join us at the 2nd Story-Level Movie Understanding & Audio Description (SLoMO) Workshop at #ECCV2026 in Malmö, Sweden!
📅 Afternoon, Sept. 8, 2026
🌐 Website: slomo-workshop.github.io/ecc…
🏆 Competition details below 👇
MovieQA track: eval.ai/web/challenges/chall…
Audio Description track: eval.ai/web/challenges/chall…
1
742
Weidi Xie retweeted
Come and see our work with @WeidiXie !!! As FIFA World Cup is coming, our SoccerMaster meets the CVPR award candidate🎉
⚽️ Introducing the first vision foundation model for this beautiful game.
🌟 CVPR 2026 Oral & Award Candidate (with @WeidiXie )
⏰ Oral: June 6, 14:00 (Four Seasons Ballroom)
📍 Poster: June 6, 16:45 (#11)
📑 SoccerMaster: A Vision Foundation Model for Soccer Understanding
Glad to share our paper [SpatialScore], as CVPR highlight.
💔Heartbroken that due to visa issues, I can’t present in person at #CVPR 2026.
🌟But still incredibly proud that our paper [SpatialScore] (with @WeidiXie) is selected as Highlight!
[CVPR 2026 Highlight] SpatialScore: Towards Comprehensive Evaluation for Spatial Intelligence
8
1,703
Come and chat with @ftm_guney ~
We show that results improve significantly after fine-tuning on real data collected this way, especially on egocentric videos.
labelling real-world videos for point tracking is cumbersome. therefore, we use synthetic training data, but this creates a gap with real-world test videos.
what if we could use existing trackers to label real-world videos? but this requires knowing which tracker to trust when.
I'm presenting our work with @gorkaydemir and @WeidiXie on learning exactly that. we show that results improve significantly after fine-tuning on real data collected this way, especially on egocentric videos.
tomorrow at the afternoon poster session (#593)
3
799
Another work on AI for Soccer ! ⚽️⚽️
Football tactics are not with a single future, In GenTac, we treat open-play as a generative multi-agent process, unifying:
⚽ multi-player trajectory forecasting
📍 tactical event recognition
Project Page: jyrao.github.io/GenTac/
Github: github.com/jyrao/GenTac
@WeidiXie
8
965
Weidi Xie retweeted
Happy to share our work: Real-World Point Tracking with Verifier-Guided Pseudo-Labeling. #CVPR2026
We improve the pseudo-label training pipeline for real-world videos using a verifier that selects the most reliable predictions across multiple trackers.
🔗kuis-ai.github.io/track_on_r
Yet, another paper on tracking, recently accepted for publication on Transactions on Pattern Analysis and Machine Intelligence (TPAMI)
It's the baseline model we used in our CVPR paper, simple and effective.
Track-On2 is finally accepted for publication in TPAMI! 🎉
congrats to @gorkaydemir and @WeidiXie!
first PAMI for all of us!
according to YÖK, I think I now officially have… 2 papers total 😂
New paper on point tracking, I personally like this paper a lot.
Trackers trained on synthetic data struggle in the real world. We introduce a meta-model that scores the reliability of multiple existing trackers frame-by-frame and selects the best predictions for fine-tuning.
Happy to share our work: Real-World Point Tracking with Verifier-Guided Pseudo-Labeling. #CVPR2026
We improve the pseudo-label training pipeline for real-world videos using a verifier that selects the most reliable predictions across multiple trackers.
🔗kuis-ai.github.io/track_on_r
Weidi Xie retweeted
Human learns from unique data -- everyone's OWN life -- but our visual representations eventually align. In our recent work "Unique Lives, Shared World" @GoogleDeepMind, we train models with "single-life" videos from distinct sources, and study their alignment and generalisation.
Weidi Xie retweeted
Welcome to join our Challenges at CVPR 2026!!! See all the details here and I'm hosting VQA challenges with $1000 prize for No.1 candidate sponsored by KNQ technology.
The SoccerNet 2026 Challenges are in full swing⚽️🔥
Our presentation video is now live : meet the task leaders and don't miss any details before the April 24, 2026 deadline.
▶️Watch it here: youtu.be/TjX789boQGM
Time to climb the leaderboard 🚀
@_CVsports #CVPR2026
Glad to share KEEP, published on Cancer Cell🥳📷
@CellPressNews
cell.com/cancer-cell/fulltex…
- KEEP is a foundation model that systematically incorporates disease knowledge into pretraining for cancer diagnosis.
- KEEP leverages a comprehensive disease knowledge graph encompassing 11,454 diseases and 139,143 attributes to reorganize millions of pathology image-text pairs into 143,000 semantically structured groups aligned with disease ontology hierarchies.
- Across 18 public benchmarks (over 14,000 whole-slide images) and 4 institutional rare cancer datasets (926 cases), KEEP consistently outperformed existing foundation models, showing substantial gains for rare subtypes.
Glad to share DeepRare, published on @Nature !
nature.com/articles/s41586-0…
Super grateful for the news & views from Prof. Timo Lassmann, it really gives the best views on DeepRare.
nature.com/articles/d41586-0…
This is the first agentic system of its kind, designed to solve the complex puzzle of rare disease diagnosis.
✅ Outperforms current best methods by ~24%
✅ 95.4% expert agreement on reasoning chains
✅ Handles clinical notes, HPO terms & genetic data
🚀 Glad to share the exciting project — SceneGen: Single-Image 3D Scene Generation in One Feedforward Pass! We explored the generation of 3D scenes with multiple assets from a single image. 🎉 ACCEPTED by 3DV 2026!!!
All resources have been open-sourced and publicly available!
📄 Paper: arxiv.org/abs/2508.15769
💻 Code: github.com/Mengmouxu/SceneGe…
🔗 Model: huggingface.co/haoningwu/Sce…
🌐 WebPage: mengmouxu.github.io/SceneGen
#3DVision #AI #GenerativeAI #ComputerVision #3DV2026 #SceneGen
Performance:
📊 SOTA Results: Tested on the 3D-FUTURE dataset, SceneGen outperforms baselines (PartCrafter, DepR) in:
✅ Geometric Accuracy (Chamfer Distance, F-score)
✅ Visual Quality (CLIP-S, DINO-S)
⚡ Efficiency: Generates a 4-asset scene in ~2 mins on a single A100 GPU.
1
269