8.15号经典模型复习笔记

文章目录

• [Deep Residual Learning for Image Recognition(CVPR2016)](#Deep Residual Learning for Image Recognition(CVPR2016))
• • 方法
• [Densely Connected Convolutional Networks（CVPR2017）](#Densely Connected Convolutional Networks（CVPR2017）)
• • 方法
• [EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks（ICML2019）](#EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks（ICML2019）)
• • 方法
• [Res2Net: A New Multi-scale Backbone Architecture](#Res2Net: A New Multi-scale Backbone Architecture)
• • 方法
• [Recurrent Residual Convolutional Neural Network based on U-Net (R2U-Net) for Medical Image Segmentation](#Recurrent Residual Convolutional Neural Network based on U-Net (R2U-Net) for Medical Image Segmentation)
• • 方法
• [Contrastive Learning of Medical Visual Representations from Paired Images and Text](#Contrastive Learning of Medical Visual Representations from Paired Images and Text)
• • 本文方法
• [RegNet: Self-Regulated Network for Image Classification](#RegNet: Self-Regulated Network for Image Classification)
• • 本文方法
• [Large-scale Robust Deep AUC Maximization: A New Surrogate Loss and Empirical Studies on Medical Image Classification（ICCV2021）](#Large-scale Robust Deep AUC Maximization: A New Surrogate Loss and Empirical Studies on Medical Image Classification（ICCV2021）)
• • 方法
• [Attention Gated Networks:Learning to Leverage Salient Regions in Medical Images](#Attention Gated Networks:Learning to Leverage Salient Regions in Medical Images)
• • 本文方法
• [Tensor Networks for Medical Image Classification（MIDL2020）](#Tensor Networks for Medical Image Classification（MIDL2020）)
• • 方法
• [SKID: Self-Supervised Learning for Knee Injury Diagnosis from MRI Data](#SKID: Self-Supervised Learning for Knee Injury Diagnosis from MRI Data)
• • 方法
• [MobileNets: Efficient Convolutional Neural Networks for Mobile Vision Applications](#MobileNets: Efficient Convolutional Neural Networks for Mobile Vision Applications)
• • 方法
• [MobileNetV2: Inverted Residuals and Linear Bottlenecks（CVPR2018）](#MobileNetV2: Inverted Residuals and Linear Bottlenecks（CVPR2018）)
• • 方法
• [VIT：An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale（ICLR2021）](#VIT：An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale（ICLR2021）)
• • 方法
• [CSPNet: A New Backbone that can Enhance Learning Capability of CNN](#CSPNet: A New Backbone that can Enhance Learning Capability of CNN)
• • 方法
• [Grad-CAM: Visual Explanations from Deep Networks via Gradient-based Localization](#Grad-CAM: Visual Explanations from Deep Networks via Gradient-based Localization)
• • 本文方法
• [SIMCLR：A Simple Framework for Contrastive Learning of Visual Representations](#SIMCLR：A Simple Framework for Contrastive Learning of Visual Representations)
• • 本文方法
• [Going Deeper with Convolutions](#Going Deeper with Convolutions)
• • 本文方法
• [Squeeze-and-Excitation Networks](#Squeeze-and-Excitation Networks)
• • 方法

Deep Residual Learning for Image Recognition(CVPR2016)

方法

resnet经典，使网络变得更深

Densely Connected Convolutional Networks（CVPR2017）

方法

每一层之间互相连接

EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks（ICML2019）

方法

相当于是在相对较小的参数下衡量最好的规模（长宽深度以及分辨率）

Res2Net: A New Multi-scale Backbone Architecture

方法

相当于是多规模

Recurrent Residual Convolutional Neural Network based on U-Net (R2U-Net) for Medical Image Segmentation

方法

我没理解错误的话相当于是保留上几步的操作的单元，类似于RNN思想

Contrastive Learning of Medical Visual Representations from Paired Images and Text

本文方法

RegNet: Self-Regulated Network for Image Classification

本文方法

可以借鉴的一个方法

Large-scale Robust Deep AUC Maximization: A New Surrogate Loss and Empirical Studies on Medical Image Classification（ICCV2021）

方法

相当于是以AUC为目标的优化，原理就不解读了，不是很简单
代码地址

Attention Gated Networks:Learning to Leverage Salient Regions in Medical Images

本文方法

相当于就是得到一个注意力系数，这个系数是关于两张特征图的

Tensor Networks for Medical Image Classification（MIDL2020）

方法

对张量进行操作的

SKID: Self-Supervised Learning for Knee Injury Diagnosis from MRI Data

方法

看代码是最好的

MobileNets: Efficient Convolutional Neural Networks for Mobile Vision Applications

方法

就是深度学分离卷积减少参数

MobileNetV2: Inverted Residuals and Linear Bottlenecks（CVPR2018）

方法

和一代相比，参数量减少，增加了残差

VIT：An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale（ICLR2021）

方法

来源于自然语言，不是很复杂，了解一下注意力计算就差不多了

CSPNet: A New Backbone that can Enhance Learning Capability of CNN

方法

看看代码就差不多了

Grad-CAM: Visual Explanations from Deep Networks via Gradient-based Localization

本文方法

相当于就是通过梯度得到可解释性的结果

SIMCLR：A Simple Framework for Contrastive Learning of Visual Representations

本文方法

两种不同的数据增强做一个对比损失

Going Deeper with Convolutions

本文方法

Squeeze-and-Excitation Networks

方法

SE模块