[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"project-82096":3},{"id":4,"name":5,"fullName":6,"owner":7,"repo":5,"description":8,"homepage":8,"htmlUrl":8,"language":9,"languages":8,"totalLinesOfCode":8,"stars":10,"forks":11,"watchers":12,"openIssues":13,"contributorsCount":14,"subscribersCount":14,"size":14,"stars1d":15,"stars7d":16,"stars30d":17,"stars90d":14,"forks30d":14,"starsTrendScore":17,"compositeScore":18,"rankGlobal":8,"rankLanguage":8,"license":19,"archived":20,"fork":20,"defaultBranch":21,"hasWiki":20,"hasPages":20,"topics":22,"createdAt":8,"pushedAt":8,"updatedAt":23,"readmeContent":24,"aiSummary":25,"trendingCount":14,"starSnapshotCount":14,"syncStatus":11,"lastSyncTime":26,"discoverSource":27},82096,"GaussianDream","TuojingAI\u002FGaussianDream","TuojingAI",null,"Python",49,2,30,1,0,6,17,18,1.43,"Apache License 2.0",false,"main",[],"2026-06-12 02:04:23","## :book: GaussianDream: A Feed-Forward 3D Gaussian World Model for Robotic Manipulation\n\n\u003Cp align=\"center\">\n  \u003Csmall>🔥 We would appreciate it if you could star GaussianDream ⭐ and share it. Thanks! 🔥\u003C\u002Fsmall>\n\u003C\u002Fp>\n\n> Zijian Zhang\u003Csup>2,3,1,*\u003C\u002Fsup>, Yuqing Jiang\u003Csup>2,3,1,*\u003C\u002Fsup>, Qian Cheng\u003Csup>4\u003C\u002Fsup>, Si Liu\u003Csup>5\u003C\u002Fsup>, Ding Zhao\u003Csup>6\u003C\u002Fsup>, Ping Luo\u003Csup>7\u003C\u002Fsup>, Weitao Zhou\u003Csup>4\u003C\u002Fsup>, Haibao Yu\u003Csup>7,1,#\u003C\u002Fsup> \u003Cbr>\n> \u003Csup>1\u003C\u002Fsup> Tuojing Intelligence, \u003Csup>2\u003C\u002Fsup> University of Chinese Academy of Sciences, \u003Csup>3\u003C\u002Fsup> Institute of Automation, Chinese Academy of Sciences \u003Cbr>\n> \u003Csup>4\u003C\u002Fsup> Tsinghua University, \u003Csup>5\u003C\u002Fsup> Beihang University, \u003Csup>6\u003C\u002Fsup> Carnegie Mellon University, \u003Csup>7\u003C\u002Fsup> The University of Hong Kong \u003Cbr>\n> \u003Csup>*\u003C\u002Fsup> Equal contribution, \u003Csup>#\u003C\u002Fsup> Corresponding author\n\n\u003Cp align=\"center\">\n\n[![Code](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FCode-GitHub-black)](https:\u002F\u002Fgithub.com\u002FTuojingAI\u002FGaussianDream) \n[![Paper](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FPaper-Coming%20Soon-red)](#release-status) \n[![Release](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FRelease-Coming%20Soon-blue)](#release-status)\n\n\u003C\u002Fp>\n\n## Introduction\n\nGaussianDream is a feed-forward 3D Gaussian world model for robotic manipulation. The core implementation lives under the `gaussiandream` Python package, while legacy checkpoint\u002Fconfig identifiers and external asset paths are kept compatible with upstream OpenPI releases.\n\n## Framework\n\n\u003Cdiv align=\"center\">\n\u003Cimg src=\"assets\u002Fillustration\u002Fframework_final.png\" \u002F>\n\u003C\u002Fdiv>\n\n## Installation\n\n```bash\ngit clone https:\u002F\u002Fgithub.com\u002FTuojingAI\u002FGaussianDream.git\ncd GaussianDream\ngit submodule update --init --recursive\nuv sync\n```\n\nThe package metadata and implementation import namespace are both `gaussiandream`.\n\nOptional cache paths:\n\n```bash\nexport GAUSSIANDREAM_DATA_HOME=\u003CCACHE_DIR>\nexport CHECKPOINT_DIR=\u003CCHECKPOINT_DIR>\nexport DATA_ROOT=\u003CDATA_ROOT>\n```\n\n`OPENPI_DATA_HOME` is still supported as a fallback for users with existing caches.\n\n## Evaluation tracks\n\nGaussianDream includes three evaluation paths:\n\n- real-robot \u002F runtime clients built around the shared policy server and `openpi-client`\n- LIBERO simulation evaluation in `examples\u002Flibero\u002F`\n- RoboCasa simulation evaluation in `examples\u002Frobocasa\u002F`\n\nStart a policy server from a checkpoint:\n\n```bash\nuv run scripts\u002Fserve_policy.py \\\n  --port 8010 \\\n  policy:checkpoint \\\n  --policy.config gaussiandream_libero \\\n  --policy.dir \u003CCHECKPOINT_DIR>\u002F\u003Cconfig>\u002F\u003Cexp>\u002F\u003Cstep>\n```\n\nFor RoboCasa checkpoints, use `--policy.config gaussiandream_robocasa` and a matching checkpoint directory.\n\n### LIBERO\n\nSee `examples\u002Flibero\u002FREADME.md` for setup and evaluation commands.\n\nTypical flow:\n\n```bash\nuv run scripts\u002Fserve_policy.py --env LIBERO\npython examples\u002Flibero\u002Fmain.py --args.task-suite-name libero_10\n```\n\n### RoboCasa\n\nSee `examples\u002Frobocasa\u002FREADME.md` for conda setup, RoboCasa assets, and rollout commands.\n\nTypical flow:\n\n```bash\nuv run scripts\u002Fserve_policy.py \\\n  --port 8010 \\\n  policy:checkpoint \\\n  --policy.config gaussiandream_robocasa \\\n  --policy.dir \u003CCHECKPOINT_DIR>\u002F\u003Cconfig>\u002F\u003Cexp>\u002F\u003Cstep>\n\npython examples\u002Frobocasa\u002Fmain.py \\\n  --host 127.0.0.1 \\\n  --port 8010 \\\n  --env-name PnPCounterToCab \\\n  --prompt \"pick and place from counter to cabinet\"\n```\n\nFor RoboCasa H50 temporal evaluation, use `examples\u002Frobocasa\u002Feval_h50_temporal.py` after installing the RoboCasa client environment.\n\n## Experiments\n\n## Release status\n\nPaper, checkpoints, datasets, and complete reproduction instructions are coming soon. Large artifacts such as datasets, checkpoints, rendered videos, logs, and experiment outputs are intentionally not tracked in git.\n\n## Contact\n\nIf you have any question, please email zijianzhang821@gmail.com.\n\n## Sincere Acknowledgement\n\nAppreciate the following works for their great contributions:\n\n- VGGT: Inspires our 3D-aware visual representation design.\n- OpenPI and π0: Serve as the foundation for our policy backbone and codebase.\n- 3D Gaussian Splatting: Inspires our 3D Gaussian world modeling design.\n- LIBERO and RoboCasa: Serve as the benchmarks for simulation training and evaluation.\n\n## Citation\n\nIf you find this work useful in your research, please cite:\n\n```bibtex\n@article{zhang2026gaussiandream,\n  title={GaussianDream: A Feed-Forward 3D Gaussian World Model for Robotic Manipulation},\n  author={Zhang, Zijian and Jiang, Yuqing and Cheng, Qian and Liu, Si and Zhao, Ding and Luo, Ping and Zhou, Weitao and Yu, Haibao},\n  year={2026}\n}\n```\n","GaussianDream 是一个用于机器人操作的前馈3D高斯世界模型。该项目的核心功能是通过Python包`gaussiandream`实现，支持与上游OpenPI版本兼容的检查点\u002F配置标识符和外部资源路径。技术特点包括提供基于共享策略服务器的真实机器人运行时客户端、LIBERO仿真评估以及RoboCasa仿真评估等三种评估路径。适合于需要在复杂环境中进行精确物体操纵任务的研究者或开发者使用，特别是在机器人学领域内探索高级感知与控制算法的应用场景中。","2026-06-11 04:07:44","CREATED_QUERY"]