Animated movies can be effortlessly understood by young minds, but appear to be challenging for video-language models, why? The key problem is the huge diversity of animated characters -- their appearance ranges from human-like faces, to cars, fish, blobs, etc.

If you are at ACM MM, don't miss it. ⚽️⚽️⚽️

Join our challenge @CVPR 2026! ⚽️⚽️🏆🏆 For $1,000 Prize with Soccer VQA task!

Yibin (@Anxiou51 ) is going to present our work ‘Learning Streaming Video Representation via Multitask Training’ at ICCV. Oral Session 3A on 8:30, Oct. 22. You can also find him at Poster Session 3 at Exhibit Hall I 224. Looking forward to seeing you there! #ICCV #ICCV2025

Happy to share our paper, ELIP: Enhanced Visual-Language Foundation Models for Image Retrieval, with Yuanpei Liu, @kaihan_x, @WeidiXie, Andrew Zisserman, which I will present at #CBMI2025(IEEE International Conference on Content-Based Multimedia Indexing) as Best Paper Candidate.

🚀 Excited to share that RadGenome-Chest CT is now published in @ScientificData! It is a comprehensive, large-scale, and fine-grained visual-language dataset based 3D CT-RATE, including: - Organ-level segmentation for 197 categories; - 665K multi-granularity grounded reports;

Some updates on point tracking, we present you: Track-On2, with - a unified memory module, - multi-scale features, and - training on longer videos. Despite being online and training only on synthetic data, Track-On2 achieves SoTA on multiple benchmarks🚀 perfect for real-time

Again, after almost two years, SAT is finally published @npjDigitalMed Thanks for all the work from reviewers and editors. 🥳 Please check the papers with changes on title and more experiments ! Large-vocabulary segmentation for medical images with text prompts.

The work towards building large-vocabulary text-prompted segmentation models for radiology, is now online at Npj Digital Medicine.

🚀 Excited to introduce our new work, Deep-DxSearch End-to-End Agentic RAG System Training for Traceable Diagnostic Reasoning ( https://t.co/hHKxBl3RLT) 🧐Objective: the goal is to transform human-prior guided LLM-powered medical RAG into end-to-end data-driven optimized agentic

After almost two years, RadFM is finally published @NatureComms Thanks for all the work from reviewers and editors. 🥳🥳 https://t.co/9PtuzmWPPK Please check the papers with changes on title: Towards generalist foundation model for radiology by leveraging web-scale 2D&3D

🚀 Excited to announce that StreamFormer is accepted to #ICCV2025 as an Oral paper! Congrats to my excellent coauthors @Anxiou51 @WeidiXie "Learning Streaming Video Representation via Multitask Training"! 📄Paper: https://t.co/h8OUpPaIVq 🌐Web: https://t.co/z2geDGRNXr

Oral@#ICCV2025 ! Only a starting point for streaming video representation learning, stay tuned, a lot more interesting work to come !

Thanks for sharing !😊

We’re hosting the 1st SLoMO Workshop at #ICCV2025 to discuss Story-Level Movie Understanding & Audio Descriptions !

Fun work on AI4Soccer ! ⚽️⚽️⚽️⚽️

For the soccer fans, don't miss this !!!!⚽️⚽️⚽️⚽️⚽️