Jiaxin Lu

🚨 Introducing HUMOTO! 🚨 Our new 4D dataset of human-object interactions with stunning details ✨, capturing daily activities from cooking 🍳 to organizing 📚. Perfect for robotics 🤖, computer vision 👁️ & animation 🎬!

Excited to share that code for the HUMOTO dataset is now available✨ github.com/Jiaxin-Lu/humo… HUMOTO is licensed to universities around the world and meets Adobe's Commercially Safe standard for research integrity! Catch us at ICCV 2025 in Hawaii🌴

I will be presenting HUMOTO in the oral session of Humanoid Agents Workshop @CVPR this afternoon from 3:30 to 4:00. Drop by and let’s talk about how we captured those “ouch I knock the cart” moment for humanoid AI.

Excited to be at @CVPR this week! I'll be presenting HUMOTO at the Humanoid Agents workshop on 6/11 afternoon. Can't wait to reconnect and meet new researchers! Let's chat about anything about 3D interaction modeling~ See you there! 🚀

I wrote a survey article (arxiv.org/abs/2410.14770) on fractured object reassembly with @jacinth_lu and a few collaborators. It covers related techniques in graphics, vision, ml and related, applications in science and medicine, and datasets.

🚨(my first) Oral paper alert! 👋 Thrilled to share that our UGG paper has been selected as ECCV 2024 Oral! 🥳 Code and models are now available—check them out! #ECCV2024 Code: github.com/Jiaxin-Lu/ugg Paper: arxiv.org/abs/2311.16917 Project page: jiaxin-lu.github.io/ugg/

👋Hope I’m not late for the party. Our paper UGG: Unified Generative Grasping has been accepted by ECCV 2024. Code and model available at github.com/Jiaxin-Lu/ugg

Head to ✈️ #CVPR2024! Looking forward to reconnecting with old friends and meeting new ones. My research focuses on object reassembly 🧩 and hand/human-object interaction👋, involving 3D reconstruction, and generation. I'd love to grab a coffee ☕️ if you're interested in my work!

@ChaerinMin2021 We hope to release the code upon acceptance of the paper :)

@jacinth_lu Hi Jiaxin, I'm very impressed by your UGG paper. Do you have any plan to open-source its code? I would like to run your code and cite the paper! Many thanks.

👋Generate dexterous grasping based on 3D input! 🤯Design 3D object based on hand pose! 💥Joint generation of 3D object and hand! Wanna put all these into one network? Check out our work UGG: Unified Generative Grasping! 🧵 👇 jiaxin-lu.github.io/ugg/

Four papers accepted to ICLR! First time I have multiple papers at a major ML conferences (my home-field is Graphics/Vision). Congratulations to my students Haitao Yang, Hanwen Jiang @hanwenjiang1 , and Siming Yan. All of them are exciting!

Excited to present my first paper @NeurIPSConf 🥳 Best swag llama🦙 and my one and only rabbit 🐰 mug from Intel🤗 Great oysters🦪 and seafoods🦞 from New Orleans😋 Great experience! Met many old friends and made a lot more!

I will be at #NeurIPS next week to present our work, Jigsaw (Wednesday, poster session 3, 210)! 🚀Apart from 3D reassembly and reconstruction, I also worked on hand-object interaction and 3D generative model.🤝🏻 Looking forward to meet friends and new friends! 🥳

Joint work with Hao Kang, Haoxiang Li, Bo Liu, Yiding Yang, @qixing_huang , Gang Hua.

🚀 UGG's standout feature? Generating objects based on given hand parameters! 💡 This not only propels object representation research in hand-object interaction tasks but opens doors to human-centric object design studies.

Our code is now open source and available at: github.com/Jiaxin-Lu/Jigs…. 🌟

It’s incredibly gratifying to see this concept validated in 3D reconstruction (a downstream task of matching). I have growing confidence that it could have a far-reaching impact on larger scale computer vision challenges. (end)

This extended wait has been a substantial obstacle for a junior student like me. I still remember the excitement of completing that paper. I strongly believe in the power of learning such universe representation under a matching perspective. 5/n

Really love what this paper did! It aligns perfectly with what I’ve been anticipating! I believe there’s more to expect based on this work. 1/n