深度学习相关论文阅读

本是古典何须时尚 2022-06-16 05:41 215阅读 0赞

## 目标检测 ##

### OverFeat ###

文章：《OverFeat: Integrated Recognition, Localization and Detection using Convolutional Networks》

### rcnn ###

代码：[https://github.com/rbgirshick/rcnn][https_github.com_rbgirshick_rcnn]，prototxt文件在`\rcnn-master\finetuning\voc_2012_prototxt`

### fast-rcnn ###

文章：《Fast R-CNN》  
代码：[https://github.com/rbgirshick/fast-rcnn][https_github.com_rbgirshick_fast-rcnn]，prototxt文件在`.\fast-rcnn-master\models\CaffeNet`

### faster-rcnn ###

文章：《Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks》  
代码：[https://github.com/rbgirshick/py-faster-rcnn][https_github.com_rbgirshick_py-faster-rcnn]，prototxt文件在`.\py-faster-rcnn-master\models\pascal_voc\VGG16\faster_rcnn_end2end`  
tensorflow的代码：[tf-faster-rcnn代码理解][tf-faster-rcnn]

### YOLO ###

文章：《You Only Look Once: Unified, Real-Time Object Detection》  
代码：[https://github.com/xingwangsfu/caffe-yolo][https_github.com_xingwangsfu_caffe-yolo]  
文章：《YOLO9000: Better, Faster, Stronger》

### SSD ###

文章：《SSD: Single Shot MultiBox Detector》  
代码：[https://github.com/weiliu89/caffe/tree/ssd][https_github.com_weiliu89_caffe_tree_ssd]  
由于SSD中需要运行`python examples/ssd/ssd_pascal.py`来生成solver.prototxt, train.prototxt, test.prototxt, deploy.prototxt。如果你没有硬件运行代码或嫌麻烦，可以在这里直接看[https://gist.github.com/JeffOwOSun/1c284e6177c38b17d83da1db7b8c8ce7][https_gist.github.com_JeffOwOSun_1c284e6177c38b17d83da1db7b8c8ce7]

### FPN ###

文章：《Feature Pyramid Networks for Object Detection》

### mask rcnn ###

文章：《Mask R-CNN》

## 语义分割 ##

### FCN ###

文章：《Fully Convolutional Networks for Semantic Segmentation》  
代码：[https://github.com/shekkizh/FCN.tensorflow][https_github.com_shekkizh_FCN.tensorflow]

## 网络结构 ##

### VGG ###

文章：《Very Deep Convolutional Networks for Large-scale image recognition》

### googleNet ###

文章：  
Inception v1:《Going Deeper with Convolutions》  
Inception v2:《Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate 》  
Inception v3:《Rethinking the Inception Architecture for Computer Vision》

### resnet ###

文章：《Deep Residual Learning for Image Recognition》  
代码：[https://github.com/KaimingHe/deep-residual-networks][https_github.com_KaimingHe_deep-residual-networks]  
补充文章：《Residual Networks Behave Like Ensembles of Relatively Shallow Networks》

## 技巧 ##

### 梯度递减优化算法综述 ###

《An overview of gradient descent optimization algorithms\*》

### dropout防止过拟合 ###

《Dropout: A Simple Way to Prevent Neural Networks from  
Overfitting》

## 其他 ##

### 《Neural Style Transfer: A Review》 ###

相关解读：[神经风格迁移研究概述：从当前研究到未来方向（附论文和代码）][Link 1]

[https_github.com_rbgirshick_rcnn]: https://github.com/rbgirshick/rcnn
[https_github.com_rbgirshick_fast-rcnn]: https://github.com/rbgirshick/fast-rcnn
[https_github.com_rbgirshick_py-faster-rcnn]: https://github.com/rbgirshick/py-faster-rcnn
[tf-faster-rcnn]: http://blog.csdn.net/u013252298/article/details/68961927?utm_source=itdadao&utm_medium=referral
[https_github.com_xingwangsfu_caffe-yolo]: https://github.com/xingwangsfu/caffe-yolo
[https_github.com_weiliu89_caffe_tree_ssd]: https://github.com/weiliu89/caffe/tree/ssd
[https_gist.github.com_JeffOwOSun_1c284e6177c38b17d83da1db7b8c8ce7]: https://gist.github.com/JeffOwOSun/1c284e6177c38b17d83da1db7b8c8ce7
[https_github.com_shekkizh_FCN.tensorflow]: https://github.com/shekkizh/FCN.tensorflow
[https_github.com_KaimingHe_deep-residual-networks]: https://github.com/KaimingHe/deep-residual-networks
[Link 1]: http://mp.weixin.qq.com/s/t_jknoYuyAM9fu6CI8OdNw