Skip to content

Residual network with 18 layers based on mnist handwritten digital image dataset.

License

Notifications You must be signed in to change notification settings

Caczhtus/deeplearning-resnet-impl

Repository files navigation

基于ResNet18的数字图像检测

简介

基于『common』包下的卷积层、池化层等基础层,实现了何凯明博士在『Deep Residual Learning for Image Recognition』论文中提到的残差单元,并拼接为『ResNet18』

快速开始

  1. 下载数据集
python mnist.py
  1. 训练并测试
python unit_test.py

代码结构

.
├── common
│   ├── functions.py
│   ├── gradient.py
│   ├── __init__.py
│   ├── layers.py
│   ├── multi_layer_net_extend.py
│   ├── multi_layer_net.py
│   ├── optimizer.py
│   ├── __pycache__
│   │   ├── functions.cpython-38.pyc
│   │   ├── __init__.cpython-38.pyc
│   │   ├── layers.cpython-38.pyc
│   │   ├── optimizer.cpython-38.pyc
│   │   ├── trainer.cpython-38.pyc
│   │   └── util.cpython-38.pyc
│   ├── trainer.py
│   └── util.py
├── mnist.pkl
├── mnist.py
├── __pycache__
│   ├── mnist.cpython-38.pyc
│   ├── Residual.cpython-38.pyc
│   └── ResNet18.cpython-38.pyc
├── Residual.py
├── ResNet18.py
├── t10k-images-idx3-ubyte.gz
├── t10k-labels-idx1-ubyte.gz
├── train-images-idx3-ubyte.gz
├── train-labels-idx1-ubyte.gz
└── unit_test.py

说明

目前本代码还有很多不足,仅供学习参考。后续有能力再进行优化补全。

  • base-resnet-block
  • resnet18
  • 普通卷积层
  • 平均池化层
  • 最大池化层
  • 线性变换层
  • 训练器

参考

  1. https://arxiv.org/pdf/1512.03385.pdf
  2. ZhangXinNan/deep_learning_from_scratch
  3. 《深度学习入门——基于Python的理论与实现》作者:斋藤康毅 译者:陆宇杰

About

Residual network with 18 layers based on mnist handwritten digital image dataset.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages