{"id":44583172,"url":"https://github.com/rr-learning/disentanglement_dataset","last_synced_at":"2026-02-14T05:57:11.369Z","repository":{"id":112320588,"uuid":"190761137","full_name":"rr-learning/disentanglement_dataset","owner":"rr-learning","description":null,"archived":false,"fork":false,"pushed_at":"2025-04-17T19:18:00.000Z","size":13423,"stargazers_count":75,"open_issues_count":1,"forks_count":2,"subscribers_count":6,"default_branch":"master","last_synced_at":"2025-04-18T09:34:43.497Z","etag":null,"topics":[],"latest_commit_sha":null,"homepage":null,"language":"Python","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"cc-by-4.0","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/rr-learning.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":"LICENSE.txt","code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null}},"created_at":"2019-06-07T14:52:18.000Z","updated_at":"2025-04-17T19:18:03.000Z","dependencies_parsed_at":"2023-05-13T05:30:57.823Z","dependency_job_id":null,"html_url":"https://github.com/rr-learning/disentanglement_dataset","commit_stats":null,"previous_names":[],"tags_count":0,"template":false,"template_full_name":null,"purl":"pkg:github/rr-learning/disentanglement_dataset","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/rr-learning%2Fdisentanglement_dataset","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/rr-learning%2Fdisentanglement_dataset/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/rr-learning%2Fdisentanglement_dataset/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/rr-learning%2Fdisentanglement_dataset/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/rr-learning","download_url":"https://codeload.github.com/rr-learning/disentanglement_dataset/tar.gz/refs/heads/master","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/rr-learning%2Fdisentanglement_dataset/sbom","scorecard":null,"host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":286080680,"owners_count":29438641,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2026-02-14T05:24:35.651Z","status":"ssl_error","status_checked_at":"2026-02-14T05:24:34.830Z","response_time":53,"last_error":"SSL_read: unexpected eof while reading","robots_txt_status":"success","robots_txt_updated_at":"2025-07-24T06:49:26.215Z","robots_txt_url":"https://github.com/robots.txt","online":false,"can_crawl_api":true,"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":[],"created_at":"2026-02-14T05:57:11.185Z","updated_at":"2026-02-14T05:57:11.359Z","avatar_url":"https://github.com/rr-learning.png","language":"Python","funding_links":[],"categories":[],"sub_categories":[],"readme":"## MPI3D Disentanglement Datasets\n\n\u003cimg src=\"https://github.com/rr-learning/disentanglement_dataset/blob/master/sample_gifs/platform.jpg\" width=\"418\" height=\"280\" /\u003e \u003cimg src=\"https://github.com/rr-learning/disentanglement_dataset/blob/master/sample_gifs/platform2.jpg\" width=\"418\" height=\"281\" /\u003e\n\nMPI3D datasets have been introduced to benchmark representations learning algorithms across simulated and real-world environments. The first transfer learning results of unsupervised disentangled representations are presented in our [NeurIPS 2019 paper.](https://proceedings.neurips.cc/paper/2019/hash/d97d404b6119214e4a7018391195240a-Abstract.html)\n\n---------------------------------\n*UPDATE:* The download links have been updated.\n---------------------------------\n\nThe dataset is also used in the [NeurIPS Disentanglement Challenge.](http://www.disentanglement-challenge.com)\nIf you use this dataset in your work then kindly cite us.\n```\n@inproceedings{NEURIPS2019_d97d404b,\n author = {Gondal, Muhammad Waleed and Wuthrich, Manuel and Miladinovic, Djordje and Locatello, Francesco and Breidt, Martin and Volchkov, Valentin and Akpo, Joel and Bachem, Olivier and Sch\\\"{o}lkopf, Bernhard and Bauer, Stefan},\n booktitle = {Advances in Neural Information Processing Systems},\n editor = {H. Wallach and H. Larochelle and A. Beygelzimer and F. d\\textquotesingle Alch\\'{e}-Buc and E. Fox and R. Garnett},\n pages = {},\n publisher = {Curran Associates, Inc.},\n title = {On the Transfer of Inductive Bias from Simulation to the Real World: a New Disentanglement Dataset},\n url = {https://proceedings.neurips.cc/paper/2019/file/d97d404b6119214e4a7018391195240a-Paper.pdf},\n volume = {32},\n year = {2019}\n}\n\n```\n\n## Datasets Information\n\nThere are following four different datasets. The gifs are created using [disentanglement_lib](https://github.com/google-research/disentanglement_lib) visualization tool.\n\n### 1. Real world simple shapes (mpi3d_real).\n\n\u003cimg src=\"https://github.com/rr-learning/disentanglement_dataset/blob/master/sample_gifs/real1.gif\"/\u003e \u003cimg src=\"https://github.com/rr-learning/disentanglement_dataset/blob/master/sample_gifs/real2.gif\" /\u003e\n\n### 2. Realistic rendered images (mpi3d_realistic).\n\n\u003cimg src=\"https://github.com/rr-learning/disentanglement_dataset/blob/master/sample_gifs/realistic1.gif\" /\u003e \u003cimg src=\"https://github.com/rr-learning/disentanglement_dataset/blob/master/sample_gifs/realistic2.gif\" /\u003e\n\n### 3. Simplistic rendered images (mpi3d_toy).\n\n\u003cimg src=\"https://github.com/rr-learning/disentanglement_dataset/blob/master/sample_gifs/toy1.gif\" /\u003e\u003cimg src=\"https://github.com/rr-learning/disentanglement_dataset/blob/master/sample_gifs/toy2.gif\" /\u003e\n\n### 4. Complex real world shapes (mpi3d_complex).\n\n\u003cimg src=\"https://github.com/rr-learning/disentanglement_dataset/blob/master/sample_gifs/complex1.gif\" /\u003e\u003cimg src=\"https://github.com/rr-learning/disentanglement_dataset/blob/master/sample_gifs/complex2.gif\" /\u003e\n\nThe **first three datasets** consists of 1,036,800 images, corresponding to all the possible combinations of the following factors of variation:\n\n|Factors|Possible Values|\n|---|---|\n|object_color|white=0, green=1, red=2, blue=3, brown=4, olive=5|\n|object_shape|cone=0, cube=1, cylinder=2, hexagonal=3, pyramid=4, sphere=5|\n|object_size|small=0, large=1|\n|camera_height|top=0, center=1, bottom=2|\n|background_color|purple=0, sea green=1, salmon=2|\n|horizontal_axis|0,...,39|\n|vertical_axis|0,...,39|\n\n\nThe **real-world complex dataset** consists of 460,800 images, containing the combinations of the following factors of variations.\n\n|Factors|Possible Values|\n|---|---|\n|object_color|yellow=0, green=1, olive=2, red=3|\n|object_shape|coffee-cup=0, tennis-ball=1, croissant=2, beer-cup=3|\n|object_size|small=0, large=1|\n|camera_height|top=0, center=1, bottom=2|\n|background_color|purple=0, sea green=1, salmon=2|\n|horizontal_axis|0,...,39|\n|vertical_axis|0,...,39|\n\n\nSo far we only provide the datasets in 64x64 resolution. Higher resolution versions will be made available in the near future.\n\n## Reading the Datasets\nThe datasets are provided in the form of numpy arrays. Once the data is loaded, you can use array.reshape([6,6,2,3,3,40,40,64,64,3]) to obtain an array where the first 7 dimensions corresponds to data generative factors as in the table above and the last three to the image dimensions. Note that for real-world complex dataset you need to use array.reshape([4,4,2,3,3,40,40,64,64,3]).\n\n```\nimport numpy as np\ndata = np.load('./mpi3d_real.npz')['images']\n\n# To visualize each factor of the data independently, you can reshape \n# the array as the following.\n\ndata = data.reshape([6,6,2,3,3,40,40,64,64,3])\n\n# For real-world complex dataset use:\ndata = data.reshape([4,4,2,3,3,40,40,64,64,3])\n```\n\n## Downloads\n\nUse the following links to download the datasets. \n\n1. mpi3d_toy (simplistic rendered):  [link](https://huggingface.co/datasets/waleedgondal/mpi3d/resolve/main/mpi3d_toy.npz)\n2. mpi3d_realistic (realistic rendered): [link](https://huggingface.co/datasets/waleedgondal/mpi3d/resolve/main/mpi3d_realistic.npz)\n3. mpi3d_real (real-world images): [link](https://huggingface.co/datasets/waleedgondal/mpi3d/resolve/main/mpi3d_real.npz)\n4. mpi3d_real_complex (real-world complex shapes images) : [link](https://drive.google.com/file/d/1Tp8eTdHxgUMtsZv5uAoYAbJR1BOa_OQm/view?usp=sharing)\n\n## Feedback\nPlease send any feedback to waleed.gondal10@gmail.com\n\n## License\n\nThis work is licensed under a Creative Commons Attribution 4.0 International License (https://creativecommons.org/licenses/by/4.0/).\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Frr-learning%2Fdisentanglement_dataset","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Frr-learning%2Fdisentanglement_dataset","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Frr-learning%2Fdisentanglement_dataset/lists"}