/ResNet-PyTorch

PyTorch implements `Deep Residual Learning for Image Recognition` paper.

Primary LanguagePythonApache License 2.0Apache-2.0

ResNet-PyTorch

Overview

This repository contains an op-for-op PyTorch reimplementation of Searching for ResNet.

Table of contents

Download weights

Download datasets

Contains MNIST, CIFAR10&CIFAR100, TinyImageNet_200, MiniImageNet_1K, ImageNet_1K, Caltech101&Caltech256 and more etc.

Please refer to README.md in the data directory for the method of making a dataset.

How Test and Train

Both training and testing only need to modify the config.py file.

Test

  • line 29: model_arch_name change to resnet18.
  • line 31: model_mean_parameters change to [0.485, 0.456, 0.406].
  • line 32: model_std_parameters change to [0.229, 0.224, 0.225].
  • line 34: model_num_classes change to 1000.
  • line 36: mode change to test.
  • line 89: model_weights_path change to ./results/pretrained_models/ResNet18-ImageNet_1K-57bb63e.pth.tar.
python3 test.py

Train model

  • line 29: model_arch_name change to resnet18.
  • line 31: model_mean_parameters change to [0.485, 0.456, 0.406].
  • line 32: model_std_parameters change to [0.229, 0.224, 0.225].
  • line 34: model_num_classes change to 1000.
  • line 36: mode change to train.
  • line 50: pretrained_model_weights_path change to ./results/pretrained_models/ResNet18-ImageNet_1K-57bb63e.pth.tar.
python3 train.py

Resume train model

  • line 29: model_arch_name change to resnet18.
  • line 31: model_mean_parameters change to [0.485, 0.456, 0.406].
  • line 32: model_std_parameters change to [0.229, 0.224, 0.225].
  • line 34: model_num_classes change to 1000.
  • line 36: mode change to train.
  • line 53: resume change to ./samples/resnet18-ImageNet_1K/epoch_xxx.pth.tar.
python3 train.py

Result

Source of original paper results: https://arxiv.org/pdf/1512.03385v1.pdf)

In the following table, the top-x error value in () indicates the result of the project, and - indicates no test.

Model Dataset Top-1 error (val) Top-5 error (val)
resnet18 ImageNet_1K 27.88%(30.25%) -(10.93%)
resnet34 ImageNet_1K 25.03%(26.71%) 7.76%(8.58%)
resnet50 ImageNet_1K 22.85%(19.65%) 6.71%(4.87%)
resnet101 ImageNet_1K 21.75%(18.33%) 6.05%(4.34%)
resnet152 ImageNet_1K 21.43%(17.66%) 5.71%(4.08%) 
# Download `ResNet18-ImageNet_1K-57bb63e.pth.tar` weights to `./results/pretrained_models`
# More detail see `README.md<Download weights>`
python3 ./inference.py

Input:

Output:

Build `resnet18` model successfully.
Load `resnet18` model weights `/ResNet-PyTorch/results/pretrained_models/ResNet18-ImageNet_1K-57bb63e.pth.tar` successfully.
tench, Tinca tinca                                                          (91.46%)
barracouta, snoek                                                           (7.15%)
gar, garfish, garpike, billfish, Lepisosteus osseus                         (0.43%)
coho, cohoe, coho salmon, blue jack, silver salmon, Oncorhynchus kisutch    (0.27%)
platypus, duckbill, duckbilled platypus, duck-billed platypus, Ornithorhynchus anatinus (0.21%)

Contributing

If you find a bug, create a GitHub issue, or even better, submit a pull request. Similarly, if you have questions, simply post them as GitHub issues.

I look forward to seeing what the community does with these models!

Credit

Deep Residual Learning for Image Recognition

Kaiming He, Xiangyu Zhang, Shaoqing Ren, Jian Sun

Abstract

Deeper neural networks are more difficult to train. We present a residual learning framework to ease the training of networks that are substantially deeper than those used previously. We explicitly reformulate the layers as learning residual functions with reference to the layer inputs, instead of learning unreferenced functions. We provide comprehensive empirical evidence showing that these residual networks are easier to optimize, and can gain accuracy from considerably increased depth. On the ImageNet dataset we evaluate residual nets with a depth of up to 152 layers---8x deeper than VGG nets but still having lower complexity. An ensemble of these residual nets achieves 3.57% error on the ImageNet test set. This result won the 1st place on the ILSVRC 2015 classification task. We also present analysis on CIFAR-10 with 100 and 1000 layers. The depth of representations is of central importance for many visual recognition tasks. Solely due to our extremely deep representations, we obtain a 28% relative improvement on the COCO object detection dataset. Deep residual nets are foundations of our submissions to ILSVRC & COCO 2015 competitions, where we also won the 1st places on the tasks of ImageNet detection, ImageNet localization, COCO detection, and COCO segmentation.

[Paper]

@inproceedings{he2016deep,
  title={Deep residual learning for image recognition},
  author={He, Kaiming and Zhang, Xiangyu and Ren, Shaoqing and Sun, Jian},
  booktitle={Proceedings of the IEEE conference on computer vision and pattern recognition},
  pages={770--778},
  year={2016}
}