{"id":17313022,"url":"https://github.com/mattdl/dua","last_synced_at":"2025-07-21T09:32:32.718Z","repository":{"id":37654285,"uuid":"248028497","full_name":"Mattdl/DUA","owner":"Mattdl","description":"Source code \"Unsupervised Model Personalization while Preserving Privacy and Scalability: An Open Problem.\" @ CVPR2020","archived":false,"fork":false,"pushed_at":"2022-12-08T10:06:31.000Z","size":260,"stargazers_count":12,"open_issues_count":4,"forks_count":3,"subscribers_count":1,"default_branch":"master","last_synced_at":"2025-04-14T14:55:41.656Z","etag":null,"topics":["cvpr2020","framework","importance","personalization","privacy","scalability","security","unsupervised-learning"],"latest_commit_sha":null,"homepage":"http://openaccess.thecvf.com/content_CVPR_2020/html/De_Lange_Unsupervised_Model_Personalization_While_Preserving_Privacy_and_Scalability_An_Open_CVPR_2020_paper.html","language":"Python","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"other","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/Mattdl.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":"LICENSE","code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null}},"created_at":"2020-03-17T17:11:39.000Z","updated_at":"2023-02-13T16:12:02.000Z","dependencies_parsed_at":"2023-01-25T06:45:07.667Z","dependency_job_id":null,"html_url":"https://github.com/Mattdl/DUA","commit_stats":null,"previous_names":[],"tags_count":0,"template":false,"template_full_name":null,"purl":"pkg:github/Mattdl/DUA","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/Mattdl%2FDUA","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/Mattdl%2FDUA/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/Mattdl%2FDUA/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/Mattdl%2FDUA/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/Mattdl","download_url":"https://codeload.github.com/Mattdl/DUA/tar.gz/refs/heads/master","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/Mattdl%2FDUA/sbom","host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":266276093,"owners_count":23903981,"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":["cvpr2020","framework","importance","personalization","privacy","scalability","security","unsupervised-learning"],"created_at":"2024-10-15T12:45:25.054Z","updated_at":"2025-07-21T09:32:32.696Z","avatar_url":"https://github.com/Mattdl.png","language":"Python","funding_links":[],"categories":[],"sub_categories":[],"readme":"# Dual User-Adaptation (DUA) framework\nSource code for CVPR2020 paper [\"Unsupervised Model Personalization while Preserving Privacy and Scalability: An Open Problem.\"](http://openaccess.thecvf.com/content_CVPR_2020/html/De_Lange_Unsupervised_Model_Personalization_While_Preserving_Privacy_and_Scalability_An_Open_CVPR_2020_paper.html)\n\n\n**In short:** Personalization of models to local user images is prone to three main problems: scalability towards thousands of users, retaining user-privacy, and labeling local user data. Our Dual User-Adaptation framework (DUA) unveils a novel perspective to tackle all of these practical concerns and enables personalization on both the server and local user edge-device.\nThe code simulates the server and users, and provides 3 benchmarks to evaluate the efficacy of our DUA framework.\n\n\u003cimg src=\"img/teaser.gif\" width=\"350\" height=\"350\"\u003e\n\n**Keywords:** Model Personalization, User Adaptation, Continual Learning, Domain Adaptation, Privacy, Scalability, Unsupervised Learning\n\n\n\n\n## Running the code\nAlways execute the scripts from within the *\"exp/\"* directory.\n- *exp/demo_script.sh*: Run demo pipeline for MAS-RACL and FIM-IMM baseline.\n- *config.init*: Adapt where to store your datasets, models and results to external paths.\n- *requirements.txt*: Install the required packages for this code.\n        ```\n                pip install -r requirements.txt\n        ```\n\nTo reproduce the results from our paper:\n- *exp/exps_Scenes.sh*: Setups to reproduce results for the MIT Indoor Scenes based dataset.\n- *exp/exps_Numbers.sh*: Setups to reproduce results for the MNIST-SVHN based Numbers dataset.\n\n## Reference Results\nResults obtained in paper: average accuracy (forgetting).\n\n1. RACL results (see *exp/exps_Scenes.sh* and *exp/exps_Numbers.sh* to replicate results)\n\n    |          | Alexnet        |                 | VGG11          |                 | MLP         |\n    |----------|----------------|-----------------|----------------|-----------------|---------------|\n    | Method   | Category Prior | Transform Prior | Category Prior | Transform Prior | Numbers       |\n    | MAS-RACL | 66.97 (0.88)   | 47.04 (-0.27)   | 77.32 (0.77)   | 53.59 (-0.14)   | 84.01 (-0.22) |\n    | FIM-RACL | 67.20 (0.73)   | 47.32 (-0.51)   | 76.53 (0.68)   | 53.73 (-0.13)   | 87.83 (0.30)  |\n    | MAS-IMM  | 67.39 (0.73)   | 46.51 (-0.14)   | 76.77 (0.30)   | 53.49 (-0.17)   | 84.36 (-0.40) |\n    | FIM-IMM  | 67.42 (0.23)   | 46.68 (-0.35)   | 76.29 (0.43)   | 53.14 (0.07)    | 87.68 (0.07)  |\n\n2. AdaBN/AdaBN-S results (see *exp/exps_Scenes.sh* to replicate results)\n\n    |               |              | CatPrior      |               |               | TransPrior    |              |              |\n    |---------------|--------------|---------------|---------------|---------------|---------------|--------------|--------------|\n    |               | Method       | BN            | AdaBN         | AdaBN-S       | BN            | AdaBN        | AdaBN-S      |\n    | User-Specific | MAS-RACL     | 58.05 (2.74)  | 58.30 (2.34)  | 60.68 (2.67)  | 30.14 (2.69)  | 30.19 (2.50) | 32.82 (3.25) |\n    |               | FIM-RACL     | 59.58 (2.14)  | 59.71 (1.61)  | 62.43 (1.84)  | 32.15 (1.53)  | 32.04 (1.33) | 34.80 (2.13) |\n    |               | Task Experts | 80.78 (5.61)  | n/a           | n/a           | 68.22 (11.35) | n/a          | n/a          |\n    | User-Agnostic | MAS-IMM      | 55.55 (2.69)  | 55.89 (2.69)  | 58.87 (2.81)  | 29.36 (2.63)  | 29.15 (2.45) | 31.73 (3.22) |\n    |               | FIM-IMM      | 61.50 (-0.03) | 61.35 (-0.46) | 63.99 (-0.16) | 32.08 (1.32)  | 31.86 (1.21) | 34.48 (2.05) |\n    |               | MAS          | 65.58 (3.96)  | 64.15 (4.04)  | 67.10 (4.66)  | 37.32 (2.64)  | 35.64 (2.88) | 40.51 (2.69) |\n    |               | EWC          | 66.20 (2.88)  | 64.03 (3.43)  | 67.54 (3.90)  | 37.16 (2.85)  | 35.44 (3.12) | 40.05 (3.18) |\n    |               | LWF          | 70.76 (0.73)  | 70.37 (0.43)  | 72.73 (1.03)  | 40.22 (0.43)  | 39.51 (0.12) | 43.07 (0.52) |\n    |               | Joint        | 75.75 (n/a)   | 72.13 (n/a)   | 76.39 (n/a)   | 46.53 (n/a)   | 41.18 (n/a)  | 48.50 (n/a)  |\n    \n## Citing and License\nUsing this code for your research? Consider citing our work:\n```\n@InProceedings{Lange_2020_CVPR,\n        author = {Lange, Matthias De and Jia, Xu and Parisot, Sarah and Leonardis, Ales and Slabaugh, Gregory and Tuytelaars, Tinne},\n        title = {Unsupervised Model Personalization While Preserving Privacy and Scalability: An Open Problem},\n        booktitle = {The IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},\n        month = {June},\n        year = {2020}\n}\n```\n\nThis source code is released under a Attribution-NonCommercial-ShareAlike 4.0 International\nlicense, hence free to use for research purposes! Find out more about it in the [LICENSE file](LICENSE).\n\nCopyright by Matthias De Lange.\n\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fmattdl%2Fdua","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fmattdl%2Fdua","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fmattdl%2Fdua/lists"}