{"id":13737635,"url":"https://github.com/juhongm999/dhpf","last_synced_at":"2025-05-08T14:33:14.745Z","repository":{"id":59272551,"uuid":"281592696","full_name":"juhongm999/dhpf","owner":"juhongm999","description":"Official PyTorch Implementation of Dynamic Hyperpixel Flow, ECCV 2020","archived":false,"fork":false,"pushed_at":"2022-12-08T02:05:22.000Z","size":30,"stargazers_count":39,"open_issues_count":0,"forks_count":4,"subscribers_count":4,"default_branch":"master","last_synced_at":"2024-11-15T06:32:13.253Z","etag":null,"topics":["computer-vision"],"latest_commit_sha":null,"homepage":"","language":"Python","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/juhongm999.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":"2020-07-22T06:19:51.000Z","updated_at":"2024-05-26T06:12:21.000Z","dependencies_parsed_at":"2023-01-25T03:00:38.720Z","dependency_job_id":null,"html_url":"https://github.com/juhongm999/dhpf","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/juhongm999%2Fdhpf","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/juhongm999%2Fdhpf/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/juhongm999%2Fdhpf/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/juhongm999%2Fdhpf/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/juhongm999","download_url":"https://codeload.github.com/juhongm999/dhpf/tar.gz/refs/heads/master","host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":253085786,"owners_count":21851700,"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":["computer-vision"],"created_at":"2024-08-03T03:01:55.690Z","updated_at":"2025-05-08T14:33:11.735Z","avatar_url":"https://github.com/juhongm999.png","language":"Python","funding_links":[],"categories":["Python"],"sub_categories":[],"readme":"[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/learning-to-compose-hypercolumns-for-visual/semantic-correspondence-on-spair-71k)](https://paperswithcode.com/sota/semantic-correspondence-on-spair-71k?p=learning-to-compose-hypercolumns-for-visual)\n[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/learning-to-compose-hypercolumns-for-visual/semantic-correspondence-on-pf-pascal)](https://paperswithcode.com/sota/semantic-correspondence-on-pf-pascal?p=learning-to-compose-hypercolumns-for-visual)\n[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/learning-to-compose-hypercolumns-for-visual/semantic-correspondence-on-pf-willow)](https://paperswithcode.com/sota/semantic-correspondence-on-pf-willow?p=learning-to-compose-hypercolumns-for-visual)\n\n# Learning to Compose Hypercolumns for Visual Correspondence\nThis is the implementation of the paper \"Learning to Compose Hypercolumns for Visual Correspondence\" by J. Min, J. Lee, J. Ponce and M. Cho. Implemented on Python 3.7 and PyTorch 1.0.1.\n\n![](https://juhongm999.github.io/pic/dhpf.png)\n\nFor more information, check out project [[website](http://cvlab.postech.ac.kr/research/DHPF/)] and the paper on [[arXiv](https://arxiv.org/abs/2007.10587)].\n\n\n## Requirements\n\n- Python 3.7\n- PyTorch 1.0.1\n- tensorboard\n- scipy\n- pandas\n- requests\n- scikit-image\n\nConda environment settings:\n```bash\nconda create -n dhpf python=3.7\nconda activate dhpf\n\nconda install pytorch=1.0.1 torchvision cudatoolkit=10.0 -c pytorch\npip install tensorboardX\nconda install -c anaconda scipy\nconda install -c anaconda pandas\nconda install -c anaconda requests\nconda install -c anaconda scikit-image\nconda install -c anaconda \"pillow\u003c7\"\n```\n\n## Training\n\nTraining DHPF with \u003cb\u003estrong supervision\u003c/b\u003e (keypoint annotations) on PF-PASCAL and SPair-71k\u003c/br\u003e\n(reproducing strongly-supervised results in Tab. 1 and 2): \n```bash\npython train.py --supervision strong \\\n                --lr 0.03 \\\n                --bsz 8 \\\n                --niter 100 \\\n                --selection 0.5 \\ \n                --benchmark pfpascal \\ \n                --backbone {resnet50, resnet101}\n\npython train.py --supervision strong \\\n                --lr 0.03 \\\n                --bsz 8 \\\n                --niter 5 \\\n                --selection 0.5 \\ \n                --benchmark spair \\ \n                --backbone {resnet50, resnet101}\n```\nTraining DHPF with \u003cb\u003eweak supervision\u003c/b\u003e (image-level labels) on PF-PASCAL\u003c/br\u003e\n(reproducing weak-supervised results in Tab. 1):\n```bash\npython train.py --supervision weak \\\n                --lr 0.1 \\\n                --bsz 4 \\\n                --niter 30 \\\n                --selection 0.5 \\  \n                --benchmark pfpascal \\\n                --backbone {resnet50, resnet101}\n```\n\n## Testing\n\nWe provide trained models available on [[Google drive](https://drive.google.com/drive/folders/1aoKQlvHOb7vZIFK8pDJsQnC7SOyEjXVF?usp=sharing)].\n\nPCK @ α\u003csub\u003eimg\u003c/sub\u003e=0.1 on PF-PASCAL at different μ:\n \n| Trained models\u003cbr\u003eat differnt μ |  0.3 |  0.4 |  0.5 |  0.6 |  0.7 |  0.8 |  0.9 |   1  |\n|:--------------------------------------------------:|:----:|:----:|:----:|:----:|:----:|:----:|:----:|:----:|\n|                    [weak (res50)](https://drive.google.com/drive/folders/1WykysKyy9PAsX-DpC5UuZILokCMToJWH?usp=sharing)                    | 77.3 |  79  |  79  | 79.3 | 79.6 | 80.7 | 81.1 | 80.7 |\n|                    [weak (res101)](https://drive.google.com/drive/folders/1IjjoFgrIZzys2YDEGhLQrOg0bTG29-Pl?usp=sharing)                   | 80.3 | 81.2 | 82.1 | 80.1 | 81.7 | 80.9 | 81.3 | 81.3 |\n|                   [strong (res50)](https://drive.google.com/drive/folders/1RC9EbVhk8QOjpF3NIO-tidIsKcY399S8?usp=sharing)                   | 87.7 | 89.1 | 88.9 | 88.5 | 89.4 | 89.1 |  89  | 89.5 |\n|                   [strong (res101)](https://drive.google.com/drive/folders/1QDYOxqF-BsWKjKbwLKfbcfxaS5OHlbVT?usp=sharing)                  | 88.7 |  90  | 90.7 | 90.2 | 90.1 | 90.6 | 90.6 | 90.4 |\n\nPCK @ α\u003csub\u003eimg\u003c/sub\u003e=0.1 on SPair-71k at μ=0.5:\n\n| Trained models\u003cbr\u003eat μ=0.5 |  PCK |\n|:---------------------------------------------:|:----:|\n|                 [weak (res101)](https://drive.google.com/file/d/1uDfONwSiAzDsxW9wbhdlYKf8auqAVXoM/view?usp=sharing)                 | 27.7 |\n|                [strong (res101)](https://drive.google.com/file/d/1DnsDhttMIImAcupdjuANowlgZqVSx_5E/view?usp=sharing)                | 37.3 |\n\nReproducing results in Tab. 1, 2 and 3:\n```bash\npython test.py --backbone {resnet50, resnet101} \\\n               --benchmark {pfpascal, pfwillow, caltech, spair} \\\n               --load \"path_to_trained_model\"\n```\n    \n   \n## BibTeX\nIf you use this code for your research, please consider citing:\n````BibTeX\n@InProceedings{min2020dhpf, \n   title={Learning to Compose Hypercolumns for Visual Correspondence},\n   author={Juhong Min and Jongmin Lee and Jean Ponce and Minsu Cho},\n   booktitle={ECCV},\n   year={2020}\n}\n````\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fjuhongm999%2Fdhpf","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fjuhongm999%2Fdhpf","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fjuhongm999%2Fdhpf/lists"}