[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"project-1006":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":17,"stars7d":18,"stars30d":19,"stars90d":16,"forks30d":16,"starsTrendScore":20,"compositeScore":21,"rankGlobal":10,"rankLanguage":10,"license":22,"archived":23,"fork":23,"defaultBranch":24,"hasWiki":25,"hasPages":23,"topics":26,"createdAt":10,"pushedAt":10,"updatedAt":30,"readmeContent":31,"aiSummary":32,"trendingCount":16,"starSnapshotCount":16,"syncStatus":33,"lastSyncTime":34,"discoverSource":35},1006,"Scal3R","zju3dv\u002FScal3R","zju3dv","[CVPR 2026 (Highlight)] Scal3R: Scalable Test-Time Training for Large-Scale 3D Reconstruction ","https:\u002F\u002Fzju3dv.github.io\u002Fscal3r",null,"Python",482,36,10,5,0,4,6,50,12,4.7,"Other",false,"main",true,[27,28,29],"3d-reconstruction","depth-estimation","large-scale","2026-06-12 02:00:21","\u003Cdiv align=\"center\">\n\n\u003Ch1>Scal3R: Scalable Test-Time Training for Large-Scale 3D Reconstruction\u003C\u002Fh1>\n\n### CVPR 2026 Highlight\n\n\u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fxbillowy\">Tao Xie\u003C\u002Fa>\u003Csup>1,2\u003C\u002Fsup>,\n\u003Ca href=\"https:\u002F\u002Fgithub.com\u002FPeiPei233\u002F\">Peishan Yang\u003C\u002Fa>\u003Csup>1\u003C\u002Fsup>,\n\u003Ca href=\"https:\u002F\u002Fkrahets.com\u002F\">Yudong Jin\u003C\u002Fa>\u003Csup>1\u003C\u002Fsup>,\nYingfeng Cai\u003Csup>2\u003C\u002Fsup>,\n\u003Ca href=\"https:\u002F\u002Fyvanyin.xyz\u002F\">Wei Yin\u003C\u002Fa>\u003Csup>2\u003C\u002Fsup>,\nWeiqiang Ren\u003Csup>2\u003C\u002Fsup>,\nQian Zhang\u003Csup>2\u003C\u002Fsup>,\n\u003Cbr>\nWei Hua\u003Csup>3\u003C\u002Fsup>,\n\u003Ca href=\"https:\u002F\u002Fpengsida.net\u002F\">Sida Peng\u003C\u002Fa>\u003Csup>1\u003C\u002Fsup>,\n\u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fxy-guo\">Xiaoyang Guo\u003C\u002Fa>\u003Csup>2†\u003C\u002Fsup>,\n\u003Ca href=\"https:\u002F\u002Fxzhou.me\u002F\">Xiaowei Zhou\u003C\u002Fa>\u003Csup>1†\u003C\u002Fsup>\n\u003Cbr>\n\n[![arXiv](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FarXiv-Paper-b31b1b?logo=arxiv&logoColor=b31b1b)](https:\u002F\u002Farxiv.org\u002Fabs\u002F2604.08542)\n[![Safari](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FProject-Page-green?logo=safari&logoColor=fff)](https:\u002F\u002Fzju3dv.github.io\u002Fscal3r)\n[![Hugging Face](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FHuggingFace-Checkpoint-yellow?logo=huggingface&logoColor=yellow)](https:\u002F\u002Fhuggingface.co\u002Fxbillowy\u002FScal3R)\n\n\u003C\u002Fdiv>\n\n***News***\n\n- 2026-04-23: Release point cloud and camera pose visualization tools.\n- 2026-04-17: Inference acceleration is enabled.\n- 2026-04-10: The inference code is released.\n- 2026-04-10: [Scal3R](https:\u002F\u002Fzju3dv.github.io\u002Fscal3r\u002F) has been selected as a highlight paper for CVPR 2026.\n\n\u003Cp align=\"center\">\n  \u003Cimg src=\"assets\u002Fteaser.gif\" alt=\"Scal3R default teaser\" width=\"100%\">\n\u003C\u002Fp>\n\n## Installation\n\nUse the automated installation script:\n\n```bash\nbash scripts\u002Finstall.sh\n```\n\nThe script creates or reuses a conda environment named `scal3r`, installs the core dependencies from `requirements.txt`, and installs Scal3R in editable mode. By default it uses `uv pip` inside that conda environment, with a plain `pip` fallback available.\n\nThis release currently includes inference only; evaluation and benchmark code are not part of the public package yet.\n\nFor detailed installation instructions and PyTorch\u002FCUDA guidance, see [docs\u002Finstall.md](docs\u002Finstall.md).\n\nDownload the required checkpoints to `data\u002Fcheckpoints\u002F`:\n\n```bash\nmkdir -p data\u002Fcheckpoints\nhf download xbillowy\u002FScal3R scal3r.pt --repo-type model --local-dir data\u002Fcheckpoints\ncurl -L https:\u002F\u002Fgithub.com\u002Fserizba\u002Fsalad\u002Freleases\u002Fdownload\u002Fv1.0.0\u002Fdino_salad.ckpt -o data\u002Fcheckpoints\u002Fdino_salad.ckpt\n```\n\n## Usage\n\nRun inference on a folder of images:\n\n```bash\npython -m scal3r.run --input_dir \u002Fpath\u002Fto\u002Fimages\n```\n\nYou can also set an explicit tag or output directory:\n\n```bash\npython -m scal3r.run \\\n  --input_dir \u002Fpath\u002Fto\u002Fimages \\\n  --tag demo \\\n  --output_dir data\u002Fresult\u002Fcustom\u002Fdemo\n```\n\nImportant arguments:\n\n- `--config`: model config path. Defaults to `configs\u002Fmodels\u002Fscal3r.yaml`.\n- `--tag`: controls the default output directory name when `--output_dir` is not set.\n- `--block_size` and `--overlap_size`: control chunking for long-sequence inference.\n- `--save_dpt` and `--save_xyz`: control whether depth maps and point clouds are exported.\n- `--offload_batches`, `--offload_outputs`: control whether to offload batches and outputs to disk.\n\nBy default, inference results are written to `data\u002Fresult\u002Fcustom\u002F\u003Ctag>\u002F`, and runtime artifacts are written to `data\u002Fresult\u002Fcustom\u002F\u003Ctag>\u002Fruntime\u002F`. The result directory typically contains:\n\n- `mat.txt` for the predicted camera poses (camera-to-world transform matrix), each row is a raveled 4x4 matrix\n- `intri.yml` and `extri.yml` for [EasyVolcap](https:\u002F\u002Fgithub.com\u002Fzju3dv\u002FEasyVolCap) format camera parameters\n- `depths\u002F` when depth export is enabled\n- `points\u002F` when point-cloud export is enabled\n- `runtime\u002F` for runtime artifacts\n\n## TODOs\n\n- [x] TODO: Release inference code.\n- [ ] TODO: Release evaluation code along with dataset preparation scripts.\n- [ ] TODO: Provide a simple viser viewer for the inference results.\n\n## Acknowledgments\n\nThis project builds on and benefits from several excellent open-source works, especially [VGGT](https:\u002F\u002Fgithub.com\u002Ffacebookresearch\u002Fvggt), [VGGT-Long](https:\u002F\u002Fgithub.com\u002FDengKaiCQ\u002FVGGT-Long), and [LaCT](https:\u002F\u002Fgithub.com\u002Fa1600012888\u002FLaCT). We thank the authors for making their code and ideas publicly available.\n\n## Citation\n\n```bibtex\n@misc{xie2026scal3rscalabletesttimetraining,\n      title={Scal3R: Scalable Test-Time Training for Large-Scale 3D Reconstruction}, \n      author={Tao Xie and Peishan Yang and Yudong Jin and Yingfeng Cai and Wei Yin and Weiqiang Ren and Qian Zhang and Wei Hua and Sida Peng and Xiaoyang Guo and Xiaowei Zhou},\n      year={2026},\n      eprint={2604.08542},\n      archivePrefix={arXiv},\n      primaryClass={cs.CV},\n      url={https:\u002F\u002Farxiv.org\u002Fabs\u002F2604.08542}, \n}\n```\n","Scal3R 是一个用于大规模3D重建的可扩展测试时训练框架。它通过在测试阶段进行训练来提高深度估计和3D重建的精度，特别适用于处理大规模数据集。项目采用Python语言开发，具有高效的推理加速功能，并提供了点云和相机姿态可视化工具。Scal3R适合需要高质量3D重建的应用场景，如自动驾驶、虚拟现实以及建筑信息建模等领域。",2,"2026-06-11 02:41:02","CREATED_QUERY"]