{"id":16887673,"url":"https://github.com/ethanhe42/resnet-tensorflow","last_synced_at":"2025-10-07T17:14:34.346Z","repository":{"id":109158165,"uuid":"57002202","full_name":"ethanhe42/ResNet-tensorflow","owner":"ethanhe42","description":"ResNet in tensorflow for CIFAR-100, CIFAR-10","archived":false,"fork":false,"pushed_at":"2017-09-27T23:59:39.000Z","size":23213,"stargazers_count":10,"open_issues_count":0,"forks_count":5,"subscribers_count":3,"default_branch":"master","last_synced_at":"2024-04-13T16:52:35.109Z","etag":null,"topics":["convolutional-layers","cpu","deepnet","kmeans","pca","residual-neural-network","tensorflow"],"latest_commit_sha":null,"homepage":"https://arxiv.org/abs/1512.03385","language":"Jupyter Notebook","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":null,"status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/ethanhe42.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":null,"code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null}},"created_at":"2016-04-25T01:06:03.000Z","updated_at":"2022-06-07T13:41:07.000Z","dependencies_parsed_at":"2023-09-24T12:09:59.338Z","dependency_job_id":null,"html_url":"https://github.com/ethanhe42/ResNet-tensorflow","commit_stats":{"total_commits":24,"total_committers":2,"mean_commits":12.0,"dds":0.08333333333333337,"last_synced_commit":"904c6c6964c3bd56b8a7e01d8a80c1038a9dd16f"},"previous_names":["ethanhe42/resnet-tensorflow"],"tags_count":0,"template":false,"template_full_name":null,"repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/ethanhe42%2FResNet-tensorflow","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/ethanhe42%2FResNet-tensorflow/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/ethanhe42%2FResNet-tensorflow/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/ethanhe42%2FResNet-tensorflow/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/ethanhe42","download_url":"https://codeload.github.com/ethanhe42/ResNet-tensorflow/tar.gz/refs/heads/master","host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":248402541,"owners_count":21097331,"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":["convolutional-layers","cpu","deepnet","kmeans","pca","residual-neural-network","tensorflow"],"created_at":"2024-10-13T16:46:10.833Z","updated_at":"2025-10-07T17:14:29.280Z","avatar_url":"https://github.com/ethanhe42.png","language":"Jupyter Notebook","funding_links":[],"categories":[],"sub_categories":[],"readme":"# Fast ResNet for CIFAR\nThis ResNet spends 3 hours to train on a modern CPU. It can reach 63% on CIFAR100 coarse 20 classes task.   \nThis residual neural network is different from the [original paper](https://github.com/KaimingHe/deep-residual-networks) in there ways:\n- Only have 13 layers, which the original paper didn't studied.  \n- No ReLU before subsampling convolutional layer, which improve accuracy by 3%  \n- BatchNorm is done before addition, which improve accuracy a little bit.  \n\nThis ResNet-13 can't beat ResNet-18/34/150/1000 layers residual nets in the long run, however, more efficient with non-sufficient training time, interestingly.  \nDetails are shown [here](report/mp2_Yihui%20He.pdf). Archtecture shown at the bottom.  \n\n### results on CIFAR-10/CIFAR-100\nTraning 3 hours on CPU:  \n\nAcc. | CIFAR-10 | CIFAR-100 \n--- | --- | ---\n[Alexnet](https://www.tensorflow.org/versions/r0.8/tutorials/deep_cnn/index.html) | 82% | -\n[Mimic Learning](resource/do-deep-nets-really-need-to-be-deep.pdf) | - | 50%\n[2-layer NN with PCA and Kmeans](https://github.com/yihui-he/Single-Layer-neural-network-with-PCAwhitening-Kmeans) | 78% | 56%\nResNet-13 (this repo) | **84%**  | **63%** \n\n### How to run  \n`python redo.py /path/to/CIFAR100python/`  \n\n### Features  \n- [x] Output layers contain 20 labels\n- [x] Using tensorflow\n- [x] number of filters\n- [x] iterations\n- [x] learning rate\n- [x] batch size\n- [x] regularization strength\n- [x] number of layers\n- [x] optimization methods\n  - [Mimic Learning](resource/do-deep-nets-really-need-to-be-deep.pdf)  \n- [x] drop out\n- [x] initialization\n  - [LSUV init](resource/ALL%20YOU%20NEED%20IS%20A%20GOOD%20INIT.pdf)\n  - Kaiming He's initialization\n- [x] hidden neurons\n- [x] filter size of convolution layers\n- [x] filter size of pooling layers  \n\n### architecture  \n![arch](report/arch.png)\n\n\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fethanhe42%2Fresnet-tensorflow","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fethanhe42%2Fresnet-tensorflow","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fethanhe42%2Fresnet-tensorflow/lists"}