Resnet (2015)

 

 

1. Introduction

 

 

 

 

 

 

2. Realted Work

Residual Representations

 

Shortcut Connections

 

 

 

 

3. Deep Residual Learning

3.1 Residual learning

 

3.2 Identity Mapping by Shortcuts

 

3.3 Network Architectures

 

Plane Network

 

Residual Network

 

 

 

 

 

3.4 Implementation

 

 

4. Experiments

 

 

4.1 ImageNet Classification

 

 

 

Plain Network

 

 

Residual Network

 

 

Identity vs Projection Shortcuts

 

 

Deeper bottleneck Architectures

 

 

50-layer ResNet 

 

 

101-layer and 152-layer ResNets

 

Comparisons with State of the art Methods

 

 

4.2 CIFAR-10 and Analysis

 

 

4.3 Object Detection on PASCAL and MS COCO