{"id":17045122,"url":"https://github.com/haozhiqi/isonet","last_synced_at":"2025-04-12T15:22:45.613Z","repository":{"id":93013299,"uuid":"276229503","full_name":"HaozhiQi/ISONet","owner":"HaozhiQi","description":"Deep Isometric Learning for Visual Recognition (ICML 2020)","archived":false,"fork":false,"pushed_at":"2022-05-29T05:57:07.000Z","size":2007,"stargazers_count":142,"open_issues_count":0,"forks_count":21,"subscribers_count":8,"default_branch":"master","last_synced_at":"2025-03-26T10:01:29.910Z","etag":null,"topics":["deep-isometric-learning","icml","visual-recognition"],"latest_commit_sha":null,"homepage":"","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/HaozhiQi.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,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null}},"created_at":"2020-06-30T23:21:03.000Z","updated_at":"2025-01-10T00:06:31.000Z","dependencies_parsed_at":"2023-04-11T16:46:23.266Z","dependency_job_id":null,"html_url":"https://github.com/HaozhiQi/ISONet","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/HaozhiQi%2FISONet","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/HaozhiQi%2FISONet/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/HaozhiQi%2FISONet/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/HaozhiQi%2FISONet/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/HaozhiQi","download_url":"https://codeload.github.com/HaozhiQi/ISONet/tar.gz/refs/heads/master","host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":248586781,"owners_count":21129099,"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":["deep-isometric-learning","icml","visual-recognition"],"created_at":"2024-10-14T09:36:30.048Z","updated_at":"2025-04-12T15:22:45.599Z","avatar_url":"https://github.com/HaozhiQi.png","language":"Python","funding_links":[],"categories":[],"sub_categories":[],"readme":"# Deep Isometric Learning for Visual Recognition\n\nThis repository is an official PyTorch implementation of the ICML paper:\n\n\u003cb\u003eDeep Isometric Learning for Visual Recognition\u003c/b\u003e \u003cbr\u003e\n[Haozhi Qi](https://people.eecs.berkeley.edu/~hqi/), \n[Chong You](https://sites.google.com/view/cyou),\n[Xiaolong Wang](https://xiaolonw.github.io/),\n[Yi Ma](http://people.eecs.berkeley.edu/~yima/),\n[Jitendra Malik](https://people.eecs.berkeley.edu/~malik/) \u003cbr\u003e\nInternational Conference on Machine Learning (ICML), 2020 \u003cbr\u003e\n[[Project Webpage](https://haozhiqi.github.io/ISONet)], [[arXiv](https://arxiv.org/abs/2006.16992)]\n\n## Introduction\n\nIn this project, we argue the notion of isometry is a central guiding principle for training deep ConvNet. In particular, we take a minimalist approach and show that a vanilla deep Isometric Network (ISONet) (i.e., without BN and shortcut) can be trained and achieve suprisingly good accuracy. We also show that if combined with skip connections, such near isometric networks (i.e. R-ISONet) can achieve performances on par with the standard ResNet, even without normalization at all.\n\n## Main results\n\nHere we show the Top-1 Classification Accuracy on ImageNet Validation dataset:\n\n| Methods | depth18 | depth 34 | depth 50 | depth 101\n| :---: | :---: | :---: | :---: | :---:  \n| ISONet | 68.10 | 70.90 | 71.20 | 71.01\n| R-ISONet | 69.17 | 73.43 | 76.18 | 77.08\n\nFor more results and pretrained models, see [Model Zoo](MODEL_ZOO.md).\n\n## Using ISONet\n\n### Data Preparation\n\nThe ImageNet folder train/val folder should locate at ```data/ILSVRC2012/train``` and ```data/ILSVRC2012/val```, respectively.\n\n### Installation\n\nThis codebase is developed and tested with python 3.6, PyTorch 1.4, and cuda 10.1. But any version newer than that should work.\n\nHere we gave an example of installing ISONet using conda virtual environment:\n```\ngit clone https://github.com/HaozhiQi/ISONet\ncd ISONet\nconda create -y -n isonet\nconda activate isonet\n# install pytorch according to https://pytorch.org/\nconda install -y pytorch==1.4 torchvision cudatoolkit=10.1 -c pytorch\npip install yacs\n```\n\n### Evaluation\n\nYou can download the pre-trained models from the links in [Model Zoo](MODEL_ZOO.md).\n\nFor example, if we want to test the performance of R-ISONet 18, download it from model zoo, and use the following command:\n\n```\n# change config files if you are going to test other pre-trained models\npython test.py --cfg configs/IN1k-RISO18.yaml --gpus {GPU_ID} --ckpt RISO18.pt\n```\n\n### Training\n\nTo train our model from scratch, use the following command:\n```\npython train.py --cfg {CONFIG_FILE} --gpus {GPU_ID} --output {OUTPUT_NAME}\n```\n\n## Citing ISONet\n\nIf you find **ISONet** or this codebase helpful in your research, please consider citing:\n```\n@InProceedings{qi2020deep,\n  author={Qi, Haozhi and You, Chong and Wang, Xiaolong and Ma, Yi and Malik, Jitendra},\n  title={Deep Isometric Learning for Visual Recognition},\n  booktitle={ICML},\n  year={2020}\n}\n```\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fhaozhiqi%2Fisonet","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fhaozhiqi%2Fisonet","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fhaozhiqi%2Fisonet/lists"}