Introducing Chapter-Llama [#CVPR2025], a framework for 𝐯𝐢𝐝𝐞𝐨 𝐜𝐡𝐚𝐩𝐭𝐞𝐫𝐢𝐧𝐠 using Large Language Models! 🎬🦙 Check it out: 📄 Paper: arxiv.org/abs/2504.00072 🔗 Project: imagine.enpc.fr/~lucas.ventura… 💻 Code: github.com/lucas-ventura/… 🤗 Demo: huggingface.co/spaces/lucas-v…

We're hiring! IMAGINE @ École des Ponts (Paris area) is opening a 4-year "CV for X" researcher position: – competitive salary – no teaching load – starting pkg ≈ 2 PhDs – goal: impactful core AI + X (climate, biodiversity, robotics...) Apply by May 31: imagine-lab.enpc.fr/wp-content/upl…

This is a smart suggestion. If language is the issue, review in your native language and then translate. Please respect your colleagues and fellow researchers!

If you're an international PhD student at Harvard studying computer vision and your visa is cancelled, reach out to me or others in Europe. Don't despair. I'm sure we can find you a great place to carry on your research.

Capturing 3D human pose and motion has made great progress but humans interact with 3D objects. Training robots from demonstration requires that we estimate the human, the object, and the contacts in 3D. PICO is a step towards that goal. At CVPR2025.

Why does 3D human-object reconstruction fail in the wild or get limited to a few object classes? A key missing piece is accurate 3D contact. InteractVLM (#CVPR2025) uses foundational models to infer contact on humans & objects, improving reconstruction from a single image. (1/10)

If you’re not in freefall, you’re likely contacting something. Yet 3D human-object interaction (HOI) reconstruction remains underexplored. PICO (#CVPR2025) recovers humans 🏃‍♂️, objects 🏓, and their interactions 👉🍎- all in 3D, from just a single internet image. 1/11

Join us today (June 12-9:00)! Rhobin workshop is happening today #CVPR2025. 🕑 9:00-12:00 📍Room 212 🎤Great talks by our speakers Federica Bogo, Li Yi, Angela Yao, Jiajun Wu, +panel discussion! 💻 Challenge winners announcement + talks!

🥁 📢🏅 The DAMON Challenge at the 3rd RHOBIN Workshop @ #CVPR2025 was a big success! With 51 teams and 108 submissions, it was fiercely competitive — pushing performance 38% to 122% beyond the DECO baseline in the Body and Semantic Contact Estimation tracks!!

It's clear that video diffusion models know a lot about the 3D world, material properties, and lighting. The trick is to get control over these. With a tiny amount of synthetic data, we can train a video model to realistically relight a single image. This is a neat trick that

Physical intelligence for humanist robots. At Meshcapade we've built the foundational technology for the capture, generation, and understanding of human motion. This blog post explains how this enables robot learning at scale. medium.com/@black_51980/p… perceiving-systems.blog/en/news/toward…