Hongchi Xia

This project is a joint effort of Hongchi, Xuan, Max @mli0603, Qianli @qianli_m, Jiashu @JiashuXu2, Ming-Yu @liu_mingyu, Yin @YinCuiCV, Tsung-Yi @TsungYiLinCV, Wei-Chiu @weichiuma, Shenlong @ShenlongWang, Shuran @SongShuran, and Fangyin @FangyinWei. Thanks to my collaborators!

To promote community research, we pre-generate 10k scenes with SAGE and name it the SAGE-10k dataset. The dataset contains 10,000 diverse scenes spanning 50 room types and styles, along with 565K uniquely generated 3D objects. Download link: huggingface.co/datasets/nvidi…

Here we introduce SAGE: Scalable Agentic 3D Scene Generation for Embodied AI, which can generate sim-ready 3D scenes with agents following user demands at scale, ready for robotic action generation. Paper, code, and SAGE-10k dataset are all released! nvlabs.github.io/sage/

I'll be at San Diego to present our #NeurIPS2025 paper! Please drop by on Friday, Dec. 5, from 4:30 to 7:30 pm at Exhibit Hall C,D,E #4901. Happy to share and discuss with everyone there!

<HoloScene> is a joint project by me (Hongchi), Chih-Hao @ZhiHaoLin16, Hao-Yu (Max) @haoyuhsu88, Quentin, Katelyn, Michael, Benjamin, and Prof. Shenlong Wang @ShenlongWang Thanks for all your contributions!

Check more results of <HoloScene> on our webpage: Webpage: xiahongchi.github.io/HoloScene/ Paper and code could be found below: Paper: arxiv.org/pdf/2510.05560 Code: github.com/xiahongchi/Hol…

#NeurIPS2025 We're excited to introduce <HoloScene>, which can create realistic and simulation-ready worlds from videos. It can recover every rigid-body object completeness, avoid geometry interpenetration, and maintain physical stability. Project page: xiahongchi.github.io/HoloScene/

DRAWER is a joint work by me (Hongchi), Entong @EntongSu, Marius @memmelma, Arhan @prodarhan, Raymond, Numfor, Prof. Ali Farhadi, Prof. Abhishek Gupta @abhishekunique7, Prof. Shenlong Wang @ShenlongWang, and Prof. Wei-Chiu Ma @weichiuma. Thanks for all your contributions!

Check more results of DRAWER on our webpage: Webpage: drawer-art.github.io Paper and code could be found below: Paper: arxiv.org/pdf/2504.15278 Code: github.com/xiahongchi/DRA…

Glad to introduce our #CVPR2025 paper "DRAWER", allowing one to create a realistic and interactable digital twin from a video of a static scene without any interactions with the environment. It unlocks many opportunities in gaming and robotics! Webpage: drawer-art.github.io