This study note is used to record my learning content during my postgraduate study.
Not long after I entered school, I found that an up master of station B made a series of sorting out + summarizing this aspect, and uploaded the code, and it is very systematic, so I plan
to Follow the steps of this up master, learn this aspect and make a record. At the same time, I will write down some of my experience. I will continue to update the blog content in the future. I plan to put the link here, and the edited content will be placed in the blog garden.
Put the link of the up master here (click me) and the GitHub link
The general content is divided into image classification, target detection, semantic segmentation, instance segmentation and key point detection.
required environment
The projects involved in this article are all completed using Pytorch
- Anaconda3 (recommended)
- python=3.6/3.7/3.8
- pycharm (IDE)
- pytorch=1.11.0 (pip package)
- torchvision=0.12.0 (pip package)
- cudatoolkit=11.3 (when pip downloads pytorch)
conda install pytorch==1.11.0 torchvision=0.12.0 cudatoolkit=11.3 -c pytorch `
Insert a piece of code to verify whether pytorch is installed
import torch
a = torch.cuda.is_available()
print(a)
ngpu= 1
device = torch.device("cuda:0" if (torch.cuda.is_available() and ngpu > 0) else "cpu")
print(device)
print(torch.cuda.get_device_name(0))
print(torch.rand(3,3).cuda())
The result after running
image classification
Image classification uses the classification of 5 types of flowers
LeNet
1998
Paper: "Gradient-Based Learning Applied to Document Recognition"
Application: PyTorch reproduces LeNet-5 study notes
AlexNet
2012
Paper: "ImageNet Classification with Deep Convolutional Neural Networks"
Application: PyTorch reproduces AlexNet study notes
VGG
2015
Paper: "VERY DEEP CONVOLUTIONAL NETWORKS FOR LARGE-SCALE IMAGE RECOGNITION"
Application: PyTorch reproduces VGG study notes
GoogLeNet
inception-v1
2015
Paper: "Going Deeper with Convolutions"
Application: PyTorch reproduces GoogleNet study notes
inception-v2
2015
Paper: "Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift" Application : Inception-v3
will be added later : "Rethinking the Inception Architecture for Computer Vision" Application: inception-v4 2016 Paper: "Inception-v4, Inception-ResNet and the Impact of Residual Connections on Learning" Application:
ResNet
2015
Paper: "Deep Residual Learning for Image Recognition"
Application: PyTorch reproduces ResNet study notes
Note: After this blog is only used to record the progress of the completion, the study notes will not be written first
ResNeXt
2017 IEEE.CVPR
paper: "Aggregated Residual Transformations for Deep Neural Networks"
application: PyTorch reproduces ResNeXt √
------------------------------- ----------2023.1.2 (completed so far)------------------------------- ---------
MobilNet _v1
2017.4.17
论文:《MobileNets: Efficient Convolutional Neural Networks for Mobile Vision
Applications》
MobilNet_v2
2018 IEEE.CVPR
paper: "MobileNetV2: Inverted Residuals and Linear Bottleneck"
application: PyTorch reproduces MobilNet_v2√
MobilNet_v3
2019.11.20
Paper: "Searching for MobileNetV3"
application: PyTorch reproduces MobilNet_v3√
ShuffleNet_v1
2017.12.7
论文:《ShuffleNet: An Extremely Efficient Convolutional Neural Network for Mobile
Devices》
ShuffleNet_v2
2018.7.30
Paper: "ShuffleNet V2: Practical Guidelines for Efficient CNN Architecture Design"
application: PyTorch reproduces ShuffleNet_v2√
Transformers
2017.12.6
Paper: "Attention Is All You Nee"
EfficientNet_v1
2020.9.11
Paper: "EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks"
Application: PyTorch reproduces EfficientNet_v1√
EfficientNet_v2
2021.6.23
Paper: "EfficientNetV2: Smaller Models and Faster Training"
Application: PyTorch reproduces EfficientNet_v2√
RepVGG
2021.5.29
paper: "RepVGG: Making VGG-style ConvNets Great Again"
application: PyTorch reproduces RepVGG√
Vsion Transformer
2021.6.3
Paper: "AN IMAGE IS WORTH 16X16 WORDS: TRANSFORMERS FOR IMAGE RECOGNITION AT SCALE"
Application: PyTorch reproduces Vsion Transformer√
Swin Transformer
2021.8.17
Paper: "Swin Transformer: Hierarchical Vision Transformer using Shifted Windows"
Application: PyTorch reproduces Swin Transformer√
ConvNeXt
2022.5.2
paper: "A ConvNet for the 2020s"
application: PyTorch reproduces ConvNeXt√
------------2023.3.15 (completed so far)------------