[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"project-72266":3},{"id":4,"name":5,"fullName":6,"owner":7,"repo":5,"description":8,"homepage":9,"htmlUrl":10,"language":11,"languages":10,"totalLinesOfCode":10,"stars":12,"forks":13,"watchers":14,"openIssues":15,"contributorsCount":16,"subscribersCount":16,"size":16,"stars1d":16,"stars7d":16,"stars30d":17,"stars90d":16,"forks30d":16,"starsTrendScore":16,"compositeScore":18,"rankGlobal":10,"rankLanguage":10,"license":19,"archived":20,"fork":20,"defaultBranch":21,"hasWiki":22,"hasPages":20,"topics":23,"createdAt":10,"pushedAt":10,"updatedAt":27,"readmeContent":28,"aiSummary":29,"trendingCount":16,"starSnapshotCount":16,"syncStatus":30,"lastSyncTime":31,"discoverSource":32},72266,"Unique3D","AiuniAI\u002FUnique3D","AiuniAI","[NeurIPS 2024] Unique3D: High-Quality and Efficient 3D Mesh Generation from a Single Image","https:\u002F\u002Fwukailu.github.io\u002FUnique3D\u002F",null,"Python",3561,289,39,79,0,6,59.99,"MIT License",false,"main",true,[24,25,26],"3d-aigc","aigc","image-to-3d","2026-06-12 04:01:04","**[中文版本](README_zh.md)**\n\n**[日本語版](README_jp.md)**\n\n# Unique3D\nOfficial implementation of Unique3D: High-Quality and Efficient 3D Mesh Generation from a Single Image. \n\n[Kailu Wu](https:\u002F\u002Fscholar.google.com\u002Fcitations?user=VTU0gysAAAAJ&hl=zh-CN&oi=ao), [Fangfu Liu](https:\u002F\u002Fliuff19.github.io\u002F), Zhihan Cai, Runjie Yan, Hanyang Wang, Yating Hu, [Yueqi Duan](https:\u002F\u002Fduanyueqi.github.io\u002F), [Kaisheng Ma](https:\u002F\u002Fgroup.iiis.tsinghua.edu.cn\u002F~maks\u002F)\n\n## [Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2405.20343) | [Project page](https:\u002F\u002Fwukailu.github.io\u002FUnique3D\u002F) | [Huggingface Demo](https:\u002F\u002Fhuggingface.co\u002Fspaces\u002FWuvin\u002FUnique3D) | [Online Demo](https:\u002F\u002Faiuni.ai\u002F#\u002Fmodeling)\n\n* Demo inference speed: Gradio Demo > Huggingface Demo > Huggingface Demo2 > Online Demo\n\n**If the Gradio Demo is overcrowded or fails to produce stable results, you can use the Online Demo [aiuni.ai](https:\u002F\u002Fwww.aiuni.ai\u002F), which is free to try (get the registration invitation code Join Discord: https:\u002F\u002Fdiscord.gg\u002Faiuni). However, the Online Demo is slightly different from the Gradio Demo, in that the inference speed is slower, but the generation is much more stable.**\n\n\u003Cp align=\"center\">\n    \u003Cimg src=\"assets\u002Fteaser_safe.jpg\">\n\u003C\u002Fp>\n\nHigh-fidelity and diverse textured meshes generated by Unique3D from single-view wild images in 30 seconds.\n\n## More features \n\nThe repo is still being under construction, thanks for your patience. \n- [x] Upload weights.\n- [x] Local gradio demo.\n- [x] Detailed tutorial.\n- [x] Huggingface demo.\n- [ ] Detailed local demo.\n- [x] Comfyui support.\n- [x] Windows support.\n- [x] Docker support.\n- [ ] More stable reconstruction with normal.\n- [ ] Training code release.\n\n## Preparation for inference\n\n* [Detailed linux installation guide](Installation.md).\n\n### Linux System Setup.\n\nAdapted for Ubuntu 22.04.4 LTS and CUDA 12.1.\n```angular2html\nconda create -n unique3d python=3.11\nconda activate unique3d\n\npip install ninja\npip install diffusers==0.27.2\n\npip install mmcv-full -f https:\u002F\u002Fdownload.openmmlab.com\u002Fmmcv\u002Fdist\u002Fcu121\u002Ftorch2.3.1\u002Findex.html\n\npip install -r requirements.txt\n```\n\n[oak-barry](https:\u002F\u002Fgithub.com\u002Foak-barry) provide another setup script for torch210+cu121 at [here](https:\u002F\u002Fgithub.com\u002Foak-barry\u002FUnique3D).\n\n### Windows Setup.\n\n* Thank you very much `jtydhr88` for the windows installation method! See [issues\u002F15](https:\u002F\u002Fgithub.com\u002FAiuniAI\u002FUnique3D\u002Fissues\u002F15).\n\nAccording to [issues\u002F15](https:\u002F\u002Fgithub.com\u002FAiuniAI\u002FUnique3D\u002Fissues\u002F15), implemented a bat script to run the commands, so you can:\n1. Might still require Visual Studio Build Tools, you can find it from [Visual Studio Build Tools](https:\u002F\u002Fvisualstudio.microsoft.com\u002Fdownloads\u002F?q=build+tools).\n2. Create conda env and activate it\n   1. `conda create -n unique3d-py311 python=3.11`\n   2. `conda activate unique3d-py311`\n3. download [triton whl](https:\u002F\u002Fhuggingface.co\u002Fmadbuda\u002Ftriton-windows-builds\u002Fresolve\u002Fmain\u002Ftriton-2.1.0-cp311-cp311-win_amd64.whl) for py311, and put it into this project.\n4. run **install_windows_win_py311_cu121.bat**\n5. answer y while asking you uninstall onnxruntime and onnxruntime-gpu\n6. create the output folder **tmp\\gradio** under the driver root, such as F:\\tmp\\gradio for me.\n7. python app\u002Fgradio_local.py --port 7860\n\nMore details prefer to [issues\u002F15](https:\u002F\u002Fgithub.com\u002FAiuniAI\u002FUnique3D\u002Fissues\u002F15).\n\n### Interactive inference: run your local gradio demo.\n\n1. Download the weights from [huggingface spaces](https:\u002F\u002Fhuggingface.co\u002Fspaces\u002FWuvin\u002FUnique3D\u002Ftree\u002Fmain\u002Fckpt) or [Tsinghua Cloud Drive](https:\u002F\u002Fcloud.tsinghua.edu.cn\u002Fd\u002F319762ec478d46c8bdf7\u002F), and extract it to `ckpt\u002F*`.\n```\nUnique3D\n    ├──ckpt\n        ├── controlnet-tile\u002F\n        ├── image2normal\u002F\n        ├── img2mvimg\u002F\n        ├── realesrgan-x4.onnx\n        └── v1-inference.yaml\n```\n\n2. Run the interactive inference locally.\n```bash\npython app\u002Fgradio_local.py --port 7860\n```\n\n## ComfyUI Support\n\nThanks for the [ComfyUI-Unique3D](https:\u002F\u002Fgithub.com\u002Fjtydhr88\u002FComfyUI-Unique3D) implementation from [jtydhr88](https:\u002F\u002Fgithub.com\u002Fjtydhr88)!\n\n## Tips to get better results\n\n**Important: Because the mesh is normalized by the longest edge of xyz during training, it is desirable that the input image needs to contain the longest edge of the object during inference, or else you may get erroneously squashed results.**\n1. Unique3D is sensitive to the facing direction of input images. Due to the distribution of the training data, orthographic front-facing images with a rest pose always lead to good reconstructions.\n2. Images with occlusions will cause worse reconstructions, since four views cannot cover the complete object. Images with fewer occlusions lead to better results.\n3. Pass an image with as high a resolution as possible to the input when resolution is a factor.\n\n## Acknowledgement\n\nWe have intensively borrowed code from the following repositories. Many thanks to the authors for sharing their code.\n- [Stable Diffusion](https:\u002F\u002Fgithub.com\u002FCompVis\u002Fstable-diffusion)\n- [Wonder3d](https:\u002F\u002Fgithub.com\u002Fxxlong0\u002FWonder3D)\n- [Zero123Plus](https:\u002F\u002Fgithub.com\u002FSUDO-AI-3D\u002Fzero123plus)\n- [Continues Remeshing](https:\u002F\u002Fgithub.com\u002FProfactor\u002Fcontinuous-remeshing)\n- [Depth from Normals](https:\u002F\u002Fgithub.com\u002FYertleTurtleGit\u002Fdepth-from-normals)\n\n## Collaborations\nOur mission is to create a 4D generative model with 3D concepts. This is just our first step, and the road ahead is still long, but we are confident. We warmly invite you to join the discussion and explore potential collaborations in any capacity. \u003Cspan style=\"color:red\">**If you're interested in connecting or partnering with us, please don't hesitate to reach out via email (wkl22@mails.tsinghua.edu.cn)**\u003C\u002Fspan>.\n\n- Follow us on twitter for the latest updates: https:\u002F\u002Fx.com\u002Faiuni_ai\n- Join AIGC 3D\u002F4D generation community on discord: https:\u002F\u002Fdiscord.gg\u002Faiuni\n- Research collaboration, please contact: ai@aiuni.ai\n\n## Citation\n\nIf you found Unique3D helpful, please cite our report:\n```bibtex\n@misc{wu2024unique3d,\n      title={Unique3D: High-Quality and Efficient 3D Mesh Generation from a Single Image}, \n      author={Kailu Wu and Fangfu Liu and Zhihan Cai and Runjie Yan and Hanyang Wang and Yating Hu and Yueqi Duan and Kaisheng Ma},\n      year={2024},\n      eprint={2405.20343},\n      archivePrefix={arXiv},\n      primaryClass={cs.CV}\n}\n```\n","Unique3D 是一个从单张图像生成高质量和高效3D网格的项目。它利用先进的深度学习技术，能够在30秒内根据输入的单视图图像生成高保真度且多样化的纹理网格。该项目支持多种操作系统包括Linux、Windows以及Docker环境，并提供了详细的安装指南和本地Gradio演示。此外，还通过Huggingface平台提供了在线演示版本，方便用户快速体验其功能。Unique3D特别适用于需要快速创建3D模型的应用场景，如游戏开发、虚拟现实内容制作以及建筑设计等领域。",2,"2026-06-11 03:41:07","high_star"]