MMSelfSup

v1.0.0 (06/04/2023)

Highlight

• Support PixMIM.
• Support DINO in projects/dino/.

New Features

• Support PixMIM (#721)
• Support DINO in projects/dino/ (#658)
• Support auto import modules from registry (#660)

Bug Fixes

• Fix registry import error of MMDet (#732)
• Fix local-rank in pytorch2.0 (#728)
• Update MAE 300e pt results (#722)
• Add missing data preprocessor in tsne configs (#715)
• Fix the bug in shape bias (#717)
• Fix T-SNE TypeError (#708)

Improvements

• Update CI (#742, #739)
• Remove file_client_args and apply new interface of fileio (#662)

Docs

• Update doc links (#738)
• Translate customize_runtime.md (#734)
• Add media links and mmpretrain announcement (#730, #693)
• Translate two docs (#725)
• Translate docs (#723)

Contributors

A total of 9 developers contributed to this release. Thanks @fangyixiao18 , @tonysy , @YuanLiuuuuuu , @vansin , @guneetmutreja , @SheffieldCao , @howayShi , @xuan07472 , @jts250

New Contributors

• @vansin made their first contribution in https://github.com/open-mmlab/mmselfsup/pull/693
• @guneetmutreja made their first contribution in https://github.com/open-mmlab/mmselfsup/pull/708
• @SheffieldCao made their first contribution in https://github.com/open-mmlab/mmselfsup/pull/723
• @howayShi made their first contribution in https://github.com/open-mmlab/mmselfsup/pull/658
• @xuan07472 made their first contribution in https://github.com/open-mmlab/mmselfsup/pull/725
• @jts250 made their first contribution in https://github.com/open-mmlab/mmselfsup/pull/734

Full Changelog: https://github.com/open-mmlab/mmselfsup/compare/v1.0.0rc6...v1.0.0

v1.0.0rc6 (10/02/2023)

Highlight

• Support MaskFeat with video dataset in projects/maskfeat_video/
• Translate documentation to Chinese.

New Features

• Support MaskFeat with video dataset in projects/maskfeat_video/ (#678)

Bug Fixes

• Fix distributed setting for shape bias (#689)
• Update link of beitv2 (#676)
• Pass param by explicitly setting location (#654)
• Update default_runtime.py (#681)
• Rename metafile.yaml to metafile.yml (#680)
• Fix bugs in configs/selfsup/eva/metafile.yaml (#669)

Improvements

• Switch default branch to 1.x (#686)
• Update pre-commit (#685)
• Deprecate the support of python 3.6 (#657)

Docs

• Translate add_transforms.md and conventions.md (#674)
• Translate classification.md, detection.md, segmentation.md (#665)
• Update link of knn script (#661)
• Translate two docs (#653)
• Translate three docs (#651)

Contributors

A total of 7 developers contributed to this release. Thanks @fangyixiao18 , @YuanLiuuuuuu , @GeoffreyChen777 , @okotaku , @Caczhtus , @123456789asdfjkl , @JetstramSam

New Contributors

• @JetstramSam made their first contribution in https://github.com/open-mmlab/mmselfsup/pull/651
• @123456789asdfjkl made their first contribution in https://github.com/open-mmlab/mmselfsup/pull/665
• @Caczhtus made their first contribution in https://github.com/open-mmlab/mmselfsup/pull/674
• @okotaku made their first contribution in https://github.com/open-mmlab/mmselfsup/pull/680
• @GeoffreyChen777 made their first contribution in https://github.com/open-mmlab/mmselfsup/pull/681

Full Changelog: https://github.com/open-mmlab/mmselfsup/compare/v1.0.0rc5...v1.0.0rc6

v1.0.0rc5 (30/12/2022)

Highlight

• Support BEiT v2, MixMIM, EVA
• Support ShapeBias for model analysis
• Add Solution of FGIA ACCV 2022 (1st Place)
• Refactor t-SNE

New Features

• Support BEiT v2 (#627)
• Support MixMIM (#626)
• Support EVA (#632)
• Support ShapeBias for model analysis (#635)
• Add convert scripts and instructions on seg and det (#621)
• Add pretraining for FGIA (#607)

Bug Fixes

• Change pseudo_collect to default_collect (#616)
• Fix the link of SimMIM 800pt 100ft (#622)
• Change map_location to cpu (#623)
• Fix import error (#631)
• Fix key error in configs (#630)
• Change np.int to int (#636)
• Fix knn multi-gpu bug (#634)

Improvements

• Refactor projects/ folder (#620)
• Refactor t-SNE (#629)
• Refactor CAE (#645)
• Refactor benchmark script and update files (#637)

Docs

• Update data_flow.md (#612)
• Update datasets.md (#633)

Contributors

A total of 6 developers contributed to this release. Thanks @fangyixiao18 , @YuanLiuuuuuu , @tonysy , @wangbo-zhao , @WasedaMagina , @ZhaoQiiii

v0.11.0 (30/12/2022)

New Features

• Support InterCLR (#609)

Bug Fixes

• Fix potential bug of hook registration (#647)
• Fix sampling_replace config kwargs bug (#646)
• Change sklearn to scikit-learn in requirements (#583)

Improvements

• Update CI check rules (#581)
• Update assignee schedule (#606)

Docs

• Add global notes and the version switcher menu (#573)

Contributors

A total of 3 developers contributed to this release. Thanks @fangyixiao18 , @Jiahao000 , @bencwallace

v1.0.0rc4 (07/12/2022)

The master branch is still 0.x version and we will checkout a new 1.x branch to release 1.x version. The two versions will be maintained simultaneously in the future.

We briefly list the major breaking changes here. Please refer to the migration guide for details and migration instructions.

Highlight

• Support BEiT and MILAN
• Support low-level reconstruction visualization

New Features

• Support BEiT (#425)
• Support MILAN (#600)
• Support low-level reconstruction visualization (#570)

Bug Fixes

• Fix registry of data preprocessor (#603)
• Fix dependence and key bug (#611)

Improvements

• Refactor file io (#582))
• Add './projects' folder and an example (#586))
• Update CI and UT (#601))

Docs

• Update readthedocs rst and menu button (#572)
• Add readthedocs algorithm pages and fix some displaying error (#599)

Contibutors

A total of 3 developers contributed to this release. Thanks @fangyixiao18 , @YuanLiuuuuuu , @soonera

v1.0.0rc3

The master branch is still 0.x version and we will checkout a new 1.x branch to release 1.x version. The two versions will be maintained simultaneously in the future.

We briefly list the major breaking changes here. Please refer to the migration guide for details and migration instructions.

Highlight

• Support MaskFeat

New Features

• Support MaskFeat (#494)
• Add Hog generator (#518)

Bug Fixes

• Fix fine-tuning config of MAE-H-448 (#509)

Improvements

• Refactor evaluation folder and related configs (#538))
• Refine configs (#547))

Docs

• Add custom dataset tutorial (#522)
• Refactor add_modules.md (#524)
• Translate some documentation to Chinese

Contributors

A total of 6 developers contributed to this release. Thanks @fangyixiao18 , @YuanLiuuuuuu , @tonysy , @Jiahao000 , @wangbo-zhao , @soonera

v0.10.1 (01/11/2022)

Improvements

• Update issue template and management file (#550, #512)
• Move res-layer to models.utils (#537)

Docs

• Add MMYOLO description for README (#541)
• Fix typo for README (#545)
• Fix lint of MaskFeat (#520)

Contributor

A total of 3 developers contributed to this release. Thanks @fangyixiao18 , @isLinXu , @nijkah

v1.0.0rc2 (12/10/2022)

The master branch is still 0.x version and we will checkout a new 1.x branch to release 1.x version. The two versions will be maintained simultaneously in the future.

We briefly list the major breaking changes here. Please refer to the migration guide for details and migration instructions.

Highlight

• Full support of MAE, SimMIM, MoCoV3.

New Features

• Full support of MAE (#483)
• Full support of SimMIM (#487)
• Full support of MoCoV3 (#496)

Bug Fixes

• Fix classification configs (#488)
• Fix MAE config name error (#498)

Improvements

• Refactor colab tutorial (#470))
• Update readthedocs requirements (#472)
• Update CI (#476)
• Refine mim_slurm_test.sh and mim_dist_test.sh for benchmarks (#477)
• Update Metafile format and content (#478)

Docs

• Add advanced_guides/engine.md (#454)
• Add advanced_guides/evaluation.md (#456)
• add advanced_guides/transforms.md (#463)
• Add dataset docs (#437)
• Refine contribution guide (#492)
• update convention (#475)

Contributors

@fangyixiao18 , @YuanLiuuuuuu , @wangbo-zhao , @soonera

v0.10.0 (30/09/2022)

Highlight

• Support MaskFeat (#485)
• Update README to announce 1.0.0rc version (#474)

New Features

• Support MaskFeat (#485)

Bug Fixes

• Fix DenseCL init weights bug (#411)
• Fix norm typo in config (#418)
• Fix read image bug (#386)

Improvements

• Change hook_cfg type access (#409)
• Support to dump training config (#410)
• Support to save MAE visualization results (#388)
• Remove default value of deprecated option (#490)

Docs

• Update the link of MAE (#497)
• Update README to announce 1.0.0rc version (#474)
• Update get_started.md (#402)
• Update model zoo (#499)

Contributors

A total of 8 developers contributed to this release. Thanks @fangyixiao18 , @YuanLiuuuuuu , @soonera , @Jason-Study , @daidaiershidi , @lorinczszabolcs , @jingt2ch , @Happylkx

We are excited to announce the release of MMSelfSup v1.0.0rc1. MMSelfSup v1.0.0rc1 is the first version of MMSelfSup 1.x, a part of the OpenMMLab 2.0 projects. The master branch is still 0.x version and we will checkout a new 1.x branch to release 1.x version. The two versions will be maintained simultaneously in the future.

We briefly list the major breaking changes here. Please refer to the migration guide for details and migration instructions.

Highlight

• Based on MMEngine and MMCV.
• Released with refactor.
  • Datasets
  • Models
  • Config
  • ...
• Refine all documents.

New Features

• Add SelfSupDataSample to unify the components' interface.
• Add SelfSupVisualizer for visualization.
• Add SelfSupDataPreprocessor for data preprocess in model.

Improvements

• Most algorithms now support non-distributed training.
• Change the interface of different data augmentation transforms to dict.
• Run classification downstream task with MMClassification.

Docs

• Refine all documents and reorganize the directory.
• Add concepts for different components.