基于pytorch构建cyclegan示例

Related tags

Deep Learningcyclegan-demo
Overview

cyclegan-demo

基于Pytorch构建CycleGAN示例

如何运行

准备数据集

将数据集整理成4个文件，分别命名为

  • trainA, trainB：训练集，A、B代表两类图片
  • testA, testB：测试集，A、B代表两类图片

例如

D:\CODE\CYCLEGAN-DEMO\DATA\SUMMER2WINTER
├─testA
├─testB
├─trainA
└─trainB

之后在main.py中将root设为数据集的路径。

参数设置

见main.py中的初始化参数

# 初始化参数
# seed: 随机种子
# root: 数据集路径
# output_model_root: 模型的输出路径
# image_size: 图片尺寸
# batch_size: 一次喂入的数据量
# lr: 学习率
# betas: 一阶和二阶动量
# epochs: 训练总次数
# historical_epochs: 历史训练次数
# - 0表示不沿用历史模型
# - >0表示对应训练次数的模型
# - -1表示最后一次训练的模型
# save_every: 保存频率
# loss_range: Loss的显示范围
seed = 123
data_root = 'D:/code/cyclegan-demo/data/summer2winter'
output_model_root = 'output/model'
image_size = 64
batch_size = 16
lr = 2e-4
betas = (.5, .999)
epochs = 100
historical_epochs = -1
save_every = 1
loss_range = 1000

安装和运行

  1. 安装依赖
pip install -r requirements.txt
  1. 打开命令行，运行Visdom
python -m visdom.server
  1. 运行主程序
python main.py

训练过程的可视化展示

访问地址http://localhost:8097即可进入Visdom可视化页面，页面中将展示：

  • A类真实图片 -【A2B生成器】 -> B类虚假图片 -【B2A生成器】 -> A类重构图片
  • B类真实图片 -【B2A生成器】 -> A类虚假图片 -【A2B生成器】 -> B类重构图片
  • 判别器A、B以及生成器的Loss曲线

一些可视化的具体用法可见Visdom的使用方法。

测试

TODO

介绍

目录结构

  • dataset.py 数据集
  • discriminator.py 判别器
  • generater.py 生成器
  • main.py 主程序
  • replay_buffer.py 缓冲区
  • resblk.py 残差块
  • util.py 工具方法

原理介绍

残差块是生成器的组成部分，其结构如下

Resblk(
  (main): Sequential(
    (0): ReflectionPad2d((1, 1, 1, 1))
    (1): Conv2d(3, 3, kernel_size=(3, 3), stride=(1, 1))
    (2): InstanceNorm2d(3, eps=1e-05, momentum=0.1, affine=False, track_running_stats=False)
    (3): ReLU(inplace=True)
    (4): ReflectionPad2d((1, 1, 1, 1))
    (5): Conv2d(3, 3, kernel_size=(3, 3), stride=(1, 1))
    (6): InstanceNorm2d(3, eps=1e-05, momentum=0.1, affine=False, track_running_stats=False)
  )
)

生成器结构如下，由于采用全卷积结构，事实上其结构与图片尺寸无关

Generater(
  (input): Sequential(
    (0): ReflectionPad2d((3, 3, 3, 3))
    (1): Conv2d(3, 64, kernel_size=(7, 7), stride=(1, 1))
    (2): InstanceNorm2d(64, eps=1e-05, momentum=0.1, affine=False, track_running_stats=False)
    (3): ReLU(inplace=True)
  )
  (downsampling): Sequential(
    (0): Conv2d(64, 128, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1))
    (1): InstanceNorm2d(128, eps=1e-05, momentum=0.1, affine=False, track_running_stats=False)
    (2): ReLU(inplace=True)
    (3): Conv2d(128, 256, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1))
    (4): InstanceNorm2d(256, eps=1e-05, momentum=0.1, affine=False, track_running_stats=False)
    (5): ReLU(inplace=True)
  )
  (resnet): Sequential(
    (0): Resblk
    (1): Resblk
    (2): Resblk
    (3): Resblk
    (4): Resblk
    (5): Resblk
    (6): Resblk
    (7): Resblk
    (8): Resblk
  )

  (upsampling): Sequential(
    (0): ConvTranspose2d(256, 128, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), output_padding=(1, 1))
    (1): InstanceNorm2d(128, eps=1e-05, momentum=0.1, affine=False, track_running_stats=False)
    (2): ReLU(inplace=True)
    (3): ConvTranspose2d(128, 64, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), output_padding=(1, 1))
    (4): InstanceNorm2d(64, eps=1e-05, momentum=0.1, affine=False, track_running_stats=False)
    (5): ReLU(inplace=True)
  )
  (output): Sequential(
    (0): ReplicationPad2d((3, 3, 3, 3))
    (1): Conv2d(64, 3, kernel_size=(7, 7), stride=(1, 1))
    (2): Tanh()
  )
)

判别器结构如下，池化层具体尺寸由图片尺寸决定，64x64的图片对应池化层为6x6

Discriminator(
  (main): Sequential(
    (0): Conv2d(3, 64, kernel_size=(4, 4), stride=(2, 2), padding=(1, 1))
    (1): LeakyReLU(negative_slope=0.2, inplace=True)
    (2): Conv2d(64, 128, kernel_size=(4, 4), stride=(2, 2), padding=(1, 1))
    (3): InstanceNorm2d(128, eps=1e-05, momentum=0.1, affine=False, track_running_stats=False)
    (4): LeakyReLU(negative_slope=0.2, inplace=True)
    (5): Conv2d(128, 256, kernel_size=(4, 4), stride=(2, 2), padding=(1, 1))
    (6): InstanceNorm2d(256, eps=1e-05, momentum=0.1, affine=False, track_running_stats=False)
    (7): LeakyReLU(negative_slope=0.2, inplace=True)
    (8): Conv2d(256, 512, kernel_size=(4, 4), stride=(1, 1), padding=(1, 1))
    (9): InstanceNorm2d(512, eps=1e-05, momentum=0.1, affine=False, track_running_stats=False)
    (10): LeakyReLU(negative_slope=0.2, inplace=True)
    (11): Conv2d(512, 1, kernel_size=(4, 4), stride=(1, 1), padding=(1, 1))
  )
  (output): Sequential(
    (0): AvgPool2d(kernel_size=torch.Size([6, 6]), stride=torch.Size([6, 6]), padding=0)
    (1): Flatten(start_dim=1, end_dim=-1)
  )
)

训练共有三个优化器，分别负责生成器、判别器A、判别器B的优化。

损失有三种类型：

  • 一致性损失：A(B)类真实图片与经生成器生成的图片的误差，该损失使得生成后的风格与原图更接近，采用L1Loss
  • 对抗损失：A(B)类图片经生成器得到B(A)类图片，再经判别器判别的错误率，采用MSELoss
  • 循环损失：A(B)类图片经生成器得到B(A)类图片，再经生成器得到A(B)类的重建图片，原图和重建图片的误差，采用L1Loss

生成器的训练过程：

  1. 将A(B)类真实图片送入生成器，得到生成的图片，计算生成图片与原图的一致性损失
  2. 将A(B)类真实图片送入生成器得到虚假图片，再送入判别器得到判别结果，计算判别结果与真实标签1的对抗损失(虚假图片应能被判别器判别为真实图片，即生成器能骗过判别器)
  3. 将A(B)类虚假图片送入生成器，得到重建图片，计算重建图片与原图的循环损失
  4. 计算、更新梯度

判别器A的训练过程：

  1. 将A类真实图片送入判别器A，得到判别结果，计算判别结果与真实标签1的对抗损失(判别器应将真实图片判别为真实)
  2. 将A类虚假图片送入判别器A，得到判别结果，计算判别结果与虚假标签0的对抗损失(判别器应将虚假图片判别为虚假)
  3. 计算、更新梯度
Owner
Koorye
学习？学个屁
Koorye
GitHub Repository
Randstad Artificial Intelligence Challenge (powered by VGEN). Soluzione proposta da Stefano Fiorucci (anakin87) - primo classificato

Randstad Artificial Intelligence Challenge (powered by VGEN) Soluzione proposta da Stefano Fiorucci (anakin87) - primo classificato Struttura director

Stefano Fiorucci 1 Nov 13, 2021
Neural-net-from-scratch - A simple Neural Network from scratch in Python using the Pymathrix library

A Simple Neural Network from scratch A Simple Neural Network from scratch in Pyt

Youssef Chafiqui 2 Jan 07, 2022
Ensembling Off-the-shelf Models for GAN Training

Vision-aided GAN video (3m) | website | paper Can the collective knowledge from a large bank of pretrained vision models be leveraged to improve GAN t

345 Dec 28, 2022
PyTorch implementation of SampleRNN: An Unconditional End-to-End Neural Audio Generation Model

samplernn-pytorch A PyTorch implementation of SampleRNN: An Unconditional End-to-End Neural Audio Generation Model. It's based on the reference implem

DeepSound 261 Dec 14, 2022
RobustART: Benchmarking Robustness on Architecture Design and Training Techniques

The first comprehensive Robustness investigation benchmark on large-scale dataset ImageNet regarding ARchitecture design and Training techniques towards diverse noises.

132 Dec 23, 2022
To propose and implement a multi-class classification approach to disaster assessment from the given data set of post-earthquake satellite imagery.

To propose and implement a multi-class classification approach to disaster assessment from the given data set of post-earthquake satellite imagery.

Kunal Wadhwa 2 Jan 05, 2022
AlgoVision - A Framework for Differentiable Algorithms and Algorithmic Supervision

NeurIPS 2021 Paper "Learning with Algorithmic Supervision via Continuous Relaxations"

Felix Petersen 76 Jan 01, 2023
Fashion Entity Classification

Fashion-Entity-Classification - Fashion-MNIST is a dataset of Zalando's article images—consisting of a training set of 60,000 examples and a test set of 10,000 examples. Each example is a 28x28 grays

ADITYA SHAH 1 Jan 04, 2022
FEMDA: Robust classification with Flexible Discriminant Analysis in heterogeneous data

FEMDA: Robust classification with Flexible Discriminant Analysis in heterogeneous data. Flexible EM-Inspired Discriminant Analysis is a robust supervised classification algorithm that performs well i

0 Sep 06, 2022
An 16kHz implementation of HiFi-GAN for soft-vc.

HiFi-GAN An 16kHz implementation of HiFi-GAN for soft-vc. Relevant links: Official HiFi-GAN repo HiFi-GAN paper Soft-VC repo Soft-VC paper Example Usa

Benjamin van Niekerk 42 Dec 27, 2022
A project which aims to protect your privacy using inexpensive hardware and easily modifiable software

Protecting your privacy using an ESP32, an IR sensor and a python script This project, which I personally call the "never-gonna-catch-me-in-the-act-ev

8 Oct 10, 2022
CSD: Consistency-based Semi-supervised learning for object Detection

CSD: Consistency-based Semi-supervised learning for object Detection (NeurIPS 2019) By Jisoo Jeong, Seungeui Lee, Jee-soo Kim, Nojun Kwak Installation

80 Dec 15, 2022
Code release for "Making a Bird AI Expert Work for You and Me".

Making-a-Bird-AI-Expert-Work-for-You-and-Me Code release for "Making a Bird AI Expert Work for You and Me". arxiv (Coming soon...) Changelog 2021/12/6

PRIS-CV: Computer Vision Group 11 Dec 11, 2022
Differentiable Prompt Makes Pre-trained Language Models Better Few-shot Learners

DART Implementation for ICLR2022 paper Differentiable Prompt Makes Pre-trained Language Models Better Few-shot Learners. Environment

ZJUNLP 83 Dec 27, 2022
Embodied Intelligence via Learning and Evolution

Embodied Intelligence via Learning and Evolution This is the code for the paper Embodied Intelligence via Learning and Evolution Agrim Gupta, Silvio S

Agrim Gupta 111 Dec 13, 2022
Official PyTorch implementation and pretrained models of the paper Self-Supervised Classification Network

Self-Classifier: Self-Supervised Classification Network Official PyTorch implementation and pretrained models of the paper Self-Supervised Classificat

Elad Amrani 24 Dec 21, 2022
The Easy-to-use Dialogue Response Selection Toolkit for Researchers

Easy-to-use toolkit for retrieval-based Chatbot Recent Activity Our released RRS corpus can be found here. Our released BERT-FP post-training checkpoi

GMFTBY 32 Nov 13, 2022
Graph InfoClust: Leveraging cluster-level node information for unsupervised graph representation learning

Graph-InfoClust-GIC [PAKDD 2021] PAKDD'21 version Graph InfoClust: Maximizing Coarse-Grain Mutual Information in Graphs Preprint version Graph InfoClu

Costas Mavromatis 21 Dec 03, 2022
Pytorch implementation of the paper "Topic Modeling Revisited: A Document Graph-based Neural Network Perspective"

Graph Neural Topic Model (GNTM) This is the pytorch implementation of the paper "Topic Modeling Revisited: A Document Graph-based Neural Network Persp

Dazhong Shen 8 Sep 14, 2022
The UI as a mobile display for OP25

OP25 Mobile Control Head A 'remote' control head that interfaces with an OP25 instance. We take advantage of some data end-points left exposed for the

Sarah Rose Giddings 13 Dec 28, 2022