{"id":13637747,"url":"https://github.com/yyliu01/ps-mt","last_synced_at":"2025-04-19T17:31:41.610Z","repository":{"id":37274664,"uuid":"473839273","full_name":"yyliu01/PS-MT","owner":"yyliu01","description":"[CVPR'22] Perturbed and Strict Mean Teachers for Semi-supervised Semantic Segmentation","archived":false,"fork":false,"pushed_at":"2023-06-27T01:45:36.000Z","size":1133,"stargazers_count":182,"open_issues_count":0,"forks_count":16,"subscribers_count":5,"default_branch":"main","last_synced_at":"2024-08-03T01:11:51.852Z","etag":null,"topics":["cvpr2022","semantic-segmentation","semi-supervised-learning","semi-supervised-segmentation"],"latest_commit_sha":null,"homepage":"https://arxiv.org/pdf/2111.12903.pdf","language":"Python","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"mit","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/yyliu01.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":"LICENSE.md","code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null}},"created_at":"2022-03-25T02:16:11.000Z","updated_at":"2024-06-24T08:55:48.000Z","dependencies_parsed_at":"2024-01-14T08:55:20.012Z","dependency_job_id":"5707ec37-719b-483b-a634-a54c675740bf","html_url":"https://github.com/yyliu01/PS-MT","commit_stats":null,"previous_names":[],"tags_count":0,"template":false,"template_full_name":null,"repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/yyliu01%2FPS-MT","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/yyliu01%2FPS-MT/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/yyliu01%2FPS-MT/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/yyliu01%2FPS-MT/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/yyliu01","download_url":"https://codeload.github.com/yyliu01/PS-MT/tar.gz/refs/heads/main","host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":223804939,"owners_count":17205824,"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":["cvpr2022","semantic-segmentation","semi-supervised-learning","semi-supervised-segmentation"],"created_at":"2024-08-02T01:00:28.764Z","updated_at":"2024-11-09T08:30:19.199Z","avatar_url":"https://github.com/yyliu01.png","language":"Python","funding_links":[],"categories":["2022"],"sub_categories":[],"readme":"# PS-MT \n\u003e **[CVPR'22]** [Perturbed and Strict Mean Teachers for Semi-supervised Semantic Segmentation](https://arxiv.org/abs/2111.12903)\n\u003e\n\u003e by Yuyuan Liu, [Yu Tian](https://yutianyt.com/), Yuanhong Chen, [Fengbei Liu](https://fbladl.github.io/), [Vasileios Belagiannis](https://campar.in.tum.de/Main/VasileiosBelagiannis) and [Gustavo Carneiro](https://cs.adelaide.edu.au/~carneiro/)\n\u003e \n\u003e Computer Vision and Pattern Recognition Conference (CVPR), 2022\n\n![image](https://user-images.githubusercontent.com/102338056/167279043-362e1405-db45-4355-b92b-0993312fe461.png)\n\n\n### Installation\nPlease install the dependencies and dataset based on this [***installation***](./docs/installation.md) document.\n\n### Getting start\nPlease follow this [***instruction***](./docs/before_start.md) document to reproduce our results.\n\n## Update\n* blender setting results in VOC12 dataset (under deeplabv3+ with resnet101)\n  \n  | Approach  | 1/16 (662)| 1/8 (1323)| 1/4 (2646)| 1/2 (5291)|\n  |:--------:\t|:-----:\t|:-----:\t|:-----:\t|:-----:\t|\n  | PS-MT  [(wandb_log)](https://wandb.ai/pyedog1976/blender-exps?workspace=user-pyedog1976)    | 78.79     | 80.29     | 80.66     | 80.87        |\n  \n  * please note that, we update the blender splits list end with an extra 0 (e.g., 6620 for 662 labels) in [the original directory](https://github.com/yyliu01/PS-MT/tree/main/VocCode/DataLoader/voc_splits). \n  * you can find the related launching scripts in [here](https://github.com/yyliu01/PS-MT/blob/main/scripts/train_voc_blender.sh). \n  * **In case you are using blender experiments (which are built on top of the high-quality labels), please compare with the results in this table**.\n  \n## Results\n### Pascal VOC12 dataset\n1. augmented set \n\n    | Backbone \t| 1/16 (662)| 1/8 (1323)| 1/4 (2646)| 1/2 (5291)|\n    |:--------:\t|:-----:\t|:-----:\t|:-----:\t|:-----:\t|\n    | 50       \t| 72.83 \t| 75.70 \t| 76.43 \t| 77.88 \t|\n    | 101      \t| 75.50 \t| 78.20 \t| 78.72 \t| 79.76 \t|\n2. high quality set (based on res101)\n\n   | 1/16 (92)| 1/8 (183)| 1/4 (366)| 1/2 (732)| full (1464)|\n   |:-----:\t|:-----:\t|:-----:\t|:-----:\t|:-----:\t|\n   | 65.80 \t| 69.58 \t| 76.57 \t| 78.42 \t|80.01|\n\n### CityScape dataset\n1. following the setting of [CAC](https://arxiv.org/pdf/2106.14133.pdf) (720x720, CE supervised loss)\n   \n    | Backbone \t| slid. eval| 1/8 (372)| 1/4 (744)| 1/2 (1488)|\n    |:--------:\t|:-----:\t|:-----:\t|:-----:\t|:-----:\t|\n    | 50       \t| ✗\t        |74.37 \t    | 75.15 \t| 76.02 \t| \n    | 50       \t| ✓\t        |75.76 \t    | 76.92 \t| 77.64 \t| \n    | 101      \t| ✓\t        |76.89\t    | 77.60 \t| 79.09 \t|       \n2. following the setting of [CPS](https://arxiv.org/pdf/2106.01226.pdf) (800x800, OHEM supervised loss)\n   \n   | Backbone \t| slid. eval| 1/8 (372)| 1/4 (744)| 1/2 (1488)|\n   |:--------:\t|:-----:\t|:-----:\t|:-----:\t|:-----:\t|\n   | 50       \t| ✓\t\t    |77.12 \t    | 78.38 \t| 79.22 \t|\n\n\n## Training details\nSome examples of training details, including:\n1) VOC12 dataset in this [wandb](https://wandb.ai/pyedog1976/PS-MT(VOC12)?workspace=user-pyedog1976) link.\n2) Cityscapes dataset in this [wandb](https://wandb.ai/pyedog1976/PS-MT(City)?workspace=user-pyedog1976) link **(w/ 1-teacher inference)**.\n\nIn details, after clicking the run (e.g., [1323_voc_rand1](https://wandb.ai/pyedog1976/PS-MT(VOC12)/runs/177s76t6?workspace=user-pyedog1976)), you can checkout:\n\n1) \u003cimg src=\"https://user-images.githubusercontent.com/102338056/167979073-1c1b3144-8a72-4d8d-9084-31d7fdab3e9b.png\" width=\"26\" height=\"22\"\u003e overall information (e.g., training command line, hardware information and training time).\n2) \u003cimg src=\"https://user-images.githubusercontent.com/102338056/167978940-8c1f3d79-d062-4e7b-b56e-30b97d273ae8.png\" width=\"26\" height=\"22\"\u003e training details (e.g., loss curves, validation results and visualization)\n3) \u003cimg src=\"https://user-images.githubusercontent.com/102338056/167979238-4847430f-aa0b-483d-b735-8a10b43293a1.png\" width=\"26\" height=\"22\"\u003e output logs (well, sometimes might crash ...)\n\n## Acknowledgement \u0026 Citation\nThe code is highly based on the [CCT](https://github.com/yassouali/CCT). Many thanks for their great work.\n\nPlease consider citing this project in your publications if it helps your research.\n```bibtex\n@article{liu2021perturbed,\n  title={Perturbed and Strict Mean Teachers for Semi-supervised Semantic Segmentation},\n  author={Liu, Yuyuan and Tian, Yu and Chen, Yuanhong and Liu, Fengbei and Belagiannis, Vasileios and Carneiro, Gustavo},\n  journal={arXiv preprint arXiv:2111.12903},\n  year={2021}\n}\n\n```\n\n#### TODO\n- [x] Code of deeplabv3+ for voc12\n- [x] Code of deeplabv3+ for cityscapes\n\n\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fyyliu01%2Fps-mt","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fyyliu01%2Fps-mt","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fyyliu01%2Fps-mt/lists"}