[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"project-80035":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":13,"contributorsCount":13,"subscribersCount":13,"size":13,"stars1d":15,"stars7d":16,"stars30d":14,"stars90d":13,"forks30d":13,"starsTrendScore":17,"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":29,"readmeContent":30,"aiSummary":31,"trendingCount":13,"starSnapshotCount":13,"syncStatus":32,"lastSyncTime":33,"discoverSource":34},80035,"d4rt","lucidrains\u002Fd4rt","lucidrains","Implementation of D4RT, Efficiently Reconstructing Dynamic Scenes, from Deepmind","",null,"Python",70,0,5,1,4,3,44.5,"MIT License",false,"main",true,[24,25,26,27,28],"4d","artificial-intelligence","attention","deep-learning","flow-matching","2026-06-12 04:01:26","\u003Cimg src=\".\u002Fd4rt.png\" width=\"400px\">\u003C\u002Fimg>\n\n## d4rt\n\nImplementation of [D4RT](https:\u002F\u002Fd4rt-paper.github.io\u002F), Efficiently Reconstructing Dynamic Scenes, by Chuhan Zhang et al. from Deepmind\n\n## install\n\n```shell\n$ pip install d4rt\n```\n\n## usage\n\n```python\nfrom torch import randn, randint\nfrom d4rt import D4RT\n\nmodel = D4RT(\n    dim = 512,\n    video_image_size = 128,\n    video_patch_size = 32,\n    video_max_time_len = 10,\n    enc_depth = 6,\n    dec_depth = 6\n)\n\nvideos = randn(2, 10, 3, 128, 128)\n\nvideo_lens = randint(2, 10, (2,)) # handle variable lengthed video, can be None for max length always\n\n# inputs\n\ncoors = randint(0, 128, (2, 5, 2))\ntime_src = randint(0, 10, (2, 5))\ntime_tgt = randint(0, 10, (2, 5))\ntime_camera = randint(0, 10, (2, 5))\n\nquery_lens = randint(1, 5, (2,)) # handle variable lengthed queries\n\n# output\n\npoints = randn(2, 5, 3)\n\nloss = model(\n    videos,\n    video_lens = video_lens,\n    coors = coors,\n    time_src = time_src,\n    time_tgt = time_tgt,\n    time_camera = time_camera,\n    query_lens = query_lens,\n    points = points,\n)\n\nloss.backward()\n\n# without giving the output, it returns the prediction\n\npred = model(\n    videos,\n    coors = coors,\n    time_src = time_src,\n    time_tgt = time_tgt,\n    time_camera = time_camera\n)\n\nassert pred.shape == (2, 5, 3)\n```\n\n## contribute\n\nJust add your code and your tests in the `tests\u002F` folder and run `pytest` in the project root\n\nVibing with attention models are welcomed\n\n## citations\n\n```bibtex\n@article{zhang2025d4rt,\n    title   = {Efficiently Reconstructing Dynamic Scenes One D4RT at a Time},\n    author  = {Zhang, Chuhan and Le Moing, Guillaume and Koppula, Skanda and Rocco, Ignacio and Momeni, Liliane and Xie, Junyu and Sun, Shuyang and Sukthankar, Rahul and Barral, Jo{\\\"e}lle K. and Hadsell, Raia and Ghahramani, Zoubin and Zisserman, Andrew and Zhang, Junlin and Sajjadi, Mehdi S. M.},\n    journal = {arXiv preprint},\n    year    = {2025}\n}\n```\n\n```bibtex\n@inproceedings{liu2026geometryaware,\n    title   = {Geometry-aware 4D Video Generation for Robot Manipulation},\n    author  = {Zeyi Liu and Shuang Li and Eric Cousineau and Siyuan Feng and Benjamin Burchfiel and Shuran Song},\n    booktitle = {The Fourteenth International Conference on Learning Representations},\n    year    = {2026},\n    url     = {https:\u002F\u002Fopenreview.net\u002Fforum?id=18gC6pZVVc}\n}\n```\n\n```bibtex\n@misc{joseph2026interpretingphysicsvideoworld,\n    title   = {Interpreting Physics in Video World Models},\n    author  = {Sonia Joseph and Quentin Garrido and Randall Balestriero and Matthew Kowal and Thomas Fel and Shahab Bakhtiari and Blake Richards and Mike Rabbat},\n    year    = {2026},\n    eprint  = {2602.07050},\n    archivePrefix = {arXiv},\n    primaryClass = {cs.CV},\n    url={https:\u002F\u002Farxiv.org\u002Fabs\u002F2602.07050},\n}\n```\n\n```bibtex\n@misc{li2025basicsletdenoisinggenerative,\n    title   = {Back to Basics: Let Denoising Generative Models Denoise},\n    author  = {Tianhong Li and Kaiming He},\n    year    = {2025},\n    eprint  = {2511.13720},\n    archivePrefix = {arXiv},\n    primaryClass = {cs.CV},\n    url     = {https:\u002F\u002Farxiv.org\u002Fabs\u002F2511.13720},\n}\n```\n\n```bibtex\n@misc{li2025basicsletdenoisinggenerative,\n    title   = {Back to Basics: Let Denoising Generative Models Denoise},\n    author  = {Tianhong Li and Kaiming He},\n    year    = {2025},\n    eprint  = {2511.13720},\n    archivePrefix = {arXiv},\n    primaryClass = {cs.CV},\n    url     = {https:\u002F\u002Farxiv.org\u002Fabs\u002F2511.13720},\n}\n```\n\n```bibtex\n@misc{charpentier2024gptbertboth,\n    title   = {GPT or BERT: why not both?},\n    author  = {Lucas Georges Gabriel Charpentier and David Samuel},\n    year    = {2024},\n    eprint  = {2410.24159},\n    archivePrefix = {arXiv},\n    primaryClass = {cs.CL},\n    url     = {https:\u002F\u002Farxiv.org\u002Fabs\u002F2410.24159},\n}\n```\n\n```bibtex\n@misc{balestriero2025lejepa,\n    title   = {LeJEPA: Provable and Scalable Self-Supervised Learning Without the Heuristics},\n    author  = {Randall Balestriero and Yann LeCun},\n    year    = {2025},\n    eprint  = {2511.08544},\n    archivePrefix = {arXiv},\n    primaryClass = {cs.LG},\n    url     = {https:\u002F\u002Farxiv.org\u002Fabs\u002F2511.08544},\n}\n```\n","d4rt 是一个实现高效动态场景重建的项目，源自 Deepmind。该项目使用 Python 开发，基于深度学习技术，特别是注意力机制和流匹配算法，能够处理具有不同长度的视频数据，并支持对视频中特定坐标点在不同时间点上的三维重建。用户可以通过简单的 API 调用来训练模型或进行预测。适用于需要从视频序列中提取并重建动态三维信息的应用场景，如机器人操控、虚拟现实环境构建等。项目采用 MIT 许可证开源，鼓励社区贡献代码与测试用例以促进其发展。",2,"2026-06-11 03:59:00","CREATED_QUERY"]