10000 GitHub - qianxihaoyue/experiment: a experiment template for all researchers to use
[go: up one dir, main page]
More Web Proxy on the site http://driver.im/
Skip to content

qianxihaoyue/experiment

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

14 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

这个项目实现了一些经典模型,方便大家学习和参考。对于刚开始入门深度学习的小伙伴有很大的帮助

UNet

论文地址:https://arxiv.org/abs/1505.04597
参考代码1:https://github.com/milesial/Pytorch-UNet
参考代码2:https://github.com/bigmb/Unet-Segmentation-Pytorch-Nest-of-Unets/tree/master
注意:
(1) 原论文的模型结构图中,卷积的padding为0,这里为了方便跳层连接,在具体实现时,padding设置为了1,使得特征层在卷积前和卷积后大小一致,仅改变通道数
(2) 原论文发表在2015年,因此BatchNorm还未发表,在本代码中,增加了BatchNorm,当batch较小时,也可以替换为InstanceNorm
(3) 在本实现中,上采样使用的转置卷积

ResNet

参考代码:https://github.com/weiaicunzai/pytorch-cifar100
参考代码:https://github.com/pytorch/vision/blob/main/torchvision/models/resnet.py

Transformer

参考博客:https://wmathor.com/index.php/archives/1438/
参考代码:https://github.com/hemingkx/ChineseNMT
参考代码:https://github.com/seanzhang-zhichen/-transformer-english2chinese-/blob/main/transformer_nmt.py

ViT

参考代码(太阳花的小绿豆,霹雳啪啦wz):https://github.com/WZMIAOMIAO/deep-learning-for-image-processing

U-Transformer

论文地址:https://arxiv.org/abs/2103.06104#
在UNet的基础上增加self-attention和cross-attention
在UNet的DouConv基础上增加一个残差连接即可
参考博客:https://www.cnblogs.com/alex-bn-lee/p/15224922.html
参考代码:https://github.com/Nishanksingla/UNet-with-ResBlock/blob/master/resnet34_unet_model.py
参考代码:https://github.com/alexbnlee/ResUNet

About

a experiment template for all researchers to use

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

0