python_developer_tools

python 开发过程中常用到的工具;包括网站开发,人工智能,文件，数据类型转换 支付接口对接，外挂，bat,sh等我在工作中常用到的开发工具

1. 制作一个可插拔的python开发工具
2. 论文复现，tricks收集

安装和使用

pip uninstall python_developer_tools
pip install git+https://github.com/carlsummer/python_developer_tools.git
from python_developer_tools import cv

---

Contents

• cv 计算机视觉

  • 基础组成成分
    • Convolution Series
      • Depthwise Separable Convolution Usage
      • MBConv
      • Involution
    • attentions注意力机制
      • Squeeze-and-Excitation Networks
    • activates 激活函数
      • DynamicReLU
    • [全连接FC]
      • SSM
  • 分类classes
    • demo
      • 训练cifar10数据集
    • 分类模型classnetwork
  • train
    • 二阶段训练
      • swa
  • utils
    • tensorboard

• files

  • common

---

---

Depthwise Separable Convolution Usage

Paper

"MobileNets: Efficient Convolutional Neural Networks for Mobile Vision Applications"

Overview

Code

感谢代码来源External-Attention-pytorch

import torch
from python_developer_tools.cv.bases.conv.DepthwiseSeparableConvolution import DepthwiseSeparableConvolution
input=torch.randn(1,3,224,224)
dsconv=DepthwiseSeparableConvolution(3,64)
out=dsconv(input)

---

---

MBConv

Paper

Rethinking model scaling for convolutional neural networks

Overview

Code

感谢代码来源External-Attention-pytorch

from python_developer_tools.cv.bases.conv.MBConv import MBConvBlock
import torch
input=torch.randn(1,3,112,112)
mbconv=MBConvBlock(ksize=3,input_filters=3,output_filters=3,image_size=112)
out=mbconv(input)
print(out.shape)

---

---

Involution

Paper

Inverting the Inherence of Convolution for Visual Recognition

Overview

Code

感谢代码来源External-Attention-pytorch

from python_developer_tools.cv.bases.conv.Involution import Involution
import torch
input=torch.randn(1,4,64,64)
involution=Involution(kernel_size=3,in_channel=4,stride=2)
out=involution(input)
print(out.shape)

---

SSM

Paper

Exploiting Featureswith Split-and-Share Module

Overview

code

from python_developer_tools.cv.bases.FC.SSM import SSM
import torch
x = torch.randn(2, 2048, 1, 1)
x = x.view(x.size(0), -1)
model = SSM()
out = model(x)
print(out.shape)

classnetwork

名称	代码实例
AlexNet	实现
DenseNet	实现
Efficientnet	实现
InceptionV1	实现
InceptionV2	实现
InceptionV3	实现
repVGGNet	实现
ResNet	实现
ResNeXt	实现
VGGNet	实现
GhostNet	实现
MixNet	实现
MobileNetV1	实现
MobileNetV2	实现
MobileNetV3	实现
MobileNetXt	实现
ShuffleNet	实现
ShuffleNetV2	实现
SqueezeNet	实现
Xception	实现

files

common

名称	功能
get_filename_suf_pix	获取路径的文件名,后缀,父路径