RuiningLi
@RayLi234
Ph.D. Student at Oxford VGG.
Introducing DSO, a method for aligning 3D generators with simulation feedback for physical soundness. At inference time, DSO-finetuned models are more likely to generate physically sound, self-supporting objects in a feed-forward manner. ⚡️Ultra-fast: no test-time optimization…
Chuanxia is amazing! He pays great attention to details and also gives students flexibility to explore. Lucky to be able to learn from him over the past 2 years! Personally, I’m bullish on foundation vision models with physical prior (whatever that means) to unlock more…
After two amazing years with @Oxford_VGG, I will be joining @NTUsg as a Nanyang Assistant Professor in Fall 2025! I’ll be leading the Physical Vision Group (physicalvision.github.io) — and we're hiring for next year!🚀 If you're passionate about vision or AI, get in touch!
#Llama4 uses inference-time temperature scaling to improve length generalization. We just released a new report on this (with @gboduljak & @jensenzhoujh)! Check it out while it's fresh: ruiningli.com/vanishing-vari… & arxiv.org/abs/2504.02827. TLDR We present a vanishing variance…
Introducing DreamHOI, a method for zero-shot human-object interaction synthesis, enabling 3D human models to interact with any object based on text, while automatically adapting to the object's specific geometry. 🕺Project: dreamhoi.github.io 🧑💻Code: github.com/hanwenzhu/drea…
Day 2 presentation. We show a new physical interaction with objects at the part level. The new puppet-master, video-based physical interaction, is also released: huggingface.co/spaces/rayli/P…. Work done by @RayLi234, w @chrirupp and Andrea @Oxford_VGG
DragAPart will be presented @eccvconf during 16:30 to 18:30 at poster #123 today! I’m not at Italy due to visa issue but the rest of the team @ChuanxiaZ @chrirupp and Andrea will be there.
Announcing DragAPart, a model that empowers image diffusion to generate nuanced part-level dynamics. Demo is on 🤗! Joint work with my amazing advisors @ChuanxiaZ, @chrirupp, and A. Vedaldi @Oxford_VGG. 📰 Paper: arxiv.org/abs/2403.15382 🌐 Page: dragapart.github.io 🧵👇
We've been working on an open-source software for 3D animal reconstruction and generation, using just images and videos available online: github.com/3DAnimals/3DAn… After all, who'd want to live in virtual worlds without animals and nature 🐾? 🦌🦒🐫🦛🦣🐎🐃🐆🐕🐿️🐇🐖
Nature presents a captivating confluence of similarity and diversity. Our new method 3D-Fauna learns a pan-category articulated 3D model of quadruped animals from Internet photos. At test time, it turns a single image into an animatable textured 3D mesh in a feed-forward pass.
🌐We released the code & data of 3D-Fauna at: github.com/3DAnimals/3DAn… 🐎🦣🐅🦒It's a **unified** codebase for several articulated animal reconstruction projects, including MagicPony (CVPR23) , 3D-Fauna (CVPR24) and Ponymation (ECCV24). Please check it out!
Nature presents a captivating confluence of similarity and diversity. Our new method 3D-Fauna learns a pan-category articulated 3D model of quadruped animals from Internet photos. At test time, it turns a single image into an animatable textured 3D mesh in a feed-forward pass.