{"id":13934960,"url":"https://github.com/google-deepmind/dsprites-dataset","last_synced_at":"2025-06-17T00:39:56.200Z","repository":{"id":43084221,"uuid":"90755551","full_name":"google-deepmind/dsprites-dataset","owner":"google-deepmind","description":"Dataset to assess the disentanglement properties of unsupervised learning methods","archived":false,"fork":false,"pushed_at":"2021-01-03T18:24:42.000Z","size":29542,"stargazers_count":499,"open_issues_count":2,"forks_count":68,"subscribers_count":13,"default_branch":"master","last_synced_at":"2025-04-15T18:49:59.042Z","etag":null,"topics":["beta-vae","dataset","disentanglement","dsprites","vae"],"latest_commit_sha":null,"homepage":"","language":"Jupyter Notebook","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"apache-2.0","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/google-deepmind.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":"CONTRIBUTING.md","funding":null,"license":"LICENSE","code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null}},"created_at":"2017-05-09T14:32:35.000Z","updated_at":"2025-04-12T14:24:42.000Z","dependencies_parsed_at":"2022-09-08T12:30:19.457Z","dependency_job_id":null,"html_url":"https://github.com/google-deepmind/dsprites-dataset","commit_stats":null,"previous_names":["google-deepmind/dsprites-dataset","deepmind/dsprites-dataset"],"tags_count":0,"template":false,"template_full_name":null,"purl":"pkg:github/google-deepmind/dsprites-dataset","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/google-deepmind%2Fdsprites-dataset","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/google-deepmind%2Fdsprites-dataset/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/google-deepmind%2Fdsprites-dataset/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/google-deepmind%2Fdsprites-dataset/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/google-deepmind","download_url":"https://codeload.github.com/google-deepmind/dsprites-dataset/tar.gz/refs/heads/master","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/google-deepmind%2Fdsprites-dataset/sbom","host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":260268635,"owners_count":22983601,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2022-07-04T15:15:14.044Z","host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":["beta-vae","dataset","disentanglement","dsprites","vae"],"created_at":"2024-08-07T23:01:19.880Z","updated_at":"2025-06-17T00:39:56.190Z","avatar_url":"https://github.com/google-deepmind.png","language":"Jupyter Notebook","funding_links":[],"categories":["Jupyter Notebook","Related Datasets Link"],"sub_categories":["**Other**"],"readme":"# dSprites - Disentanglement testing Sprites dataset\n\nThis repository contains the dSprites dataset, used to assess the\ndisentanglement properties of unsupervised learning methods.\n\nIf you use this dataset in your work, please cite it as follows:\n\n## Bibtex\n\n```\n@misc{dsprites17,\nauthor = {Loic Matthey and Irina Higgins and Demis Hassabis and Alexander Lerchner},\ntitle = {dSprites: Disentanglement testing Sprites dataset},\nhowpublished= {https://github.com/deepmind/dsprites-dataset/},\nyear = \"2017\",\n}\n```\n\n## Description\n\n![dsprite_gif](dsprites.gif)\n\ndSprites is a dataset of 2D shapes procedurally generated from 6 ground truth\nindependent latent factors. These factors are *color*, *shape*, *scale*,\n*rotation*, *x* and *y* positions of a sprite.\n\nAll possible combinations of these latents are present exactly once,\ngenerating N = 737280 total images.\n\n### Latent factor values\n\n*   Color: white\n*   Shape: square, ellipse, heart\n*   Scale: 6 values linearly spaced in [0.5, 1]\n*   Orientation: 40 values in [0, 2 pi]\n*   Position X: 32 values in [0, 1]\n*   Position Y: 32 values in [0, 1]\n\nWe varied one latent at a time (starting from Position Y, then Position X, etc),\nand sequentially stored the images in fixed order.\nHence the order along the first dimension is fixed and allows you to map back to\nthe value of the latents corresponding to that image.\n\nWe chose the latents values deliberately to have the smallest step changes\nwhile ensuring that all pixel outputs were different. No noise was added.\n\nThe data is a NPZ NumPy archive with the following fields:\n\n*   `imgs`: (737280 x 64 x 64, uint8) Images in black and white.\n*   `latents_values`: (737280 x 6, float64) Values of the latent factors.\n*   `latents_classes`: (737280 x 6, int64) Integer index of the latent factor\n    values. Useful as classification targets.\n*   `metadata`: some additional information, including the possible latent\n    values.\n\nAlternatively, a HDF5 version is also available, containing the same data,\npacked as Groups and Datasets.\n\n## Disentanglement metric\n\nThis dataset was created as a unit test of disentanglement properties of\nunsupervised models. It can be used to determine how well models recover the\nground truth latents presented above.\n\nYou find our proposed disentanglement metric assessing the disentanglement\nquality of a model (along with an example usage of this dataset) in:\n\n[Higgins, Irina, Loic Matthey, Arka Pal, Christopher Burgess, Xavier Glorot,\nMatthew Botvinick, Shakir Mohamed, and Alexander Lerchner. \"beta-VAE: Learning\nbasic visual concepts with a constrained variational framework.\" In *Proceedings\nof the International Conference on Learning Representations (ICLR).\n2017.*](https://openreview.net/forum?id=Sy2fzU9gl)\n\n## Disclaimers\n\nThis is not an official Google product.\n\nThe images were generated using the LOVE framework, which is licenced under\nzlib/libpng licence:\n\n```\nLOVE is Copyright (c) 2006-2016 LOVE Development Team\n\nThis software is provided 'as-is', without any express or implied\nwarranty. In no event will the authors be held liable for any damages\narising from the use of this software.\n\nPermission is granted to anyone to use this software for any purpose,\nincluding commercial applications, and to alter it and redistribute it\nfreely, subject to the following restrictions:\n\n1. The origin of this software must not be misrepresented; you must not\nclaim that you wrote the original software. If you use this software\nin a product, an acknowledgment in the product documentation would be\nappreciated but is not required.\n\n2. Altered source versions must be plainly marked as such, and must not be\nmisrepresented as being the original software.\n\n3. This notice may not be removed or altered from any source\ndistribution.\n```\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fgoogle-deepmind%2Fdsprites-dataset","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fgoogle-deepmind%2Fdsprites-dataset","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fgoogle-deepmind%2Fdsprites-dataset/lists"}