---
title: 109/12/03
---

## Paper
### 特徵檢測
[Look Closer to See Better: Recurrent Attention Convolutional Neural Network
for Fine-grained Image Recognition(CVPR2017)](https://openaccess.thecvf.com/content_cvpr_2017/papers/Fu_Look_Closer_to_CVPR_2017_paper.pdf)
有點像做bounding box的效果

### 改善模型方面
[Squeeze-and-Excitation Networks (2017)](https://arxiv.org/pdf/1709.01507.pdf)

將channel attention加入現有的模型，並改善了原有的效果
[CBAM: Convolutional Block Attention Module (2018)](https://arxiv.org/pdf/1807.06521v2.pdf)

提出一個新的block包含channel attention和spatial ttention
[AN IMAGE IS WORTH 16X16 WORDS:TRANSFORMERS FOR IMAGE RECOGNITION AT SCALE (2020/10/22,ICLR 2021)](https://arxiv.org/pdf/2010.11929.pdf)
NLP概念取代傳統CNN結構，包含MSA、MLP等不同的概念