【文献阅读】Perceptual Generative Adversarial Networks for Small Object Detection –CVPR-2017
来源:互联网 发布:同城国际旅行社 java 编辑:程序博客网 时间:2024/05/19 13:57
Perceptual Generative Adversarial Networks for Small Object Detection
2017CVPR 新鲜出炉的paper,这是针对small object detection的一篇文章,采用PGAN来提升small object detection任务的performance。
最近也没做object detection,只是别人推荐了这篇paper,看了摘要觉得通俗易懂就往下看了。。。最后发现还是没怎么搞懂,只是明白PGAN的模型。如果理解有误的地方,请指出。
言归正传,PGAN为什么对small object有效?具体是这样,small object 不好检测,而large object好检测,那PGAN就让generator 学习一个映射,把small object 的features 映射成 large object 的features,然后就好检测了。PGAN呢,主要就看它的generator。
传统GAN中的generator是学习从随机噪声到图像的映射,也就是generator可以把一个噪声变成图片,而PGAN的思想是让generator把small object 变成 large object,这样就有利于检测了。 来看看文章中的原话都是怎么介绍generator的:
- we address the small object detection problem by developing a single architecture that internally lifts representations of small objects to “super-resolved” ones, achieving similar characteristics as large objects
- Perceptual Generative Adversarial Network (Perceptual GAN) model that improves small object detection through narrowing representation difference of small objects from the large ones.
- generator learns to transfer perceived poor representations of the small objects to super-resolved ones
- The Perceptual GAN aims to enhance the representations of small objects to be similar to those of large object
- the generator is a deep residual based feature generative model which transforms the original poor features of small objects to highly discriminative ones by introducing fine-grained details from lower-level layers, achieving “super-resolution” on the intermediate representations
6.传统的generator G represents a generator that learns to map data z from the noise distribution pz(z) to the distribution pdata(x) over data x,而PGAN的generator中 x and z are the representations for large objects and small objects - The generator network aims to generate super-resolved representations for small objects to improve detection accurac
- the generator as a deep residual learning network that augments the representations of small objects to super-resolved ones by introducing more fine-grained details absent from the small objects through residual learning
文章在不同地方不断的重复了一个意思,就是generator学习的是一个映射,这个映射就是把假(small object)的变成真(large object)的
来看看generator长什么样子
分两个部分,这里就没看懂是什么意思了,或许和object detection有关了。最终得出的结果是Super-Resolved Features 这个就很像Large Objects Featuresle. 如图,左下角是G生成的,左上角是真实的:
讲完了generator 就到discriminator了,这里的discrimintor和传统的GAN也有不一样的地方。
在这里,加入了一个新的loss,叫做perceptual loss ,PGAN也因此而得名(我猜的,很明显嘛)这个loss我也是没看明白的地方,贴原文大家看看吧(有理解的这部分的同学,请在评论区讲一讲,供大家学习)
1. justify the detection accuracy benefiting from the generated super-resolved features with a perceptual loss
看完paper感觉作者没有很直接说提出PGAN是inspired by哪些文章~不过GAN(2014 Goodfellow)
- 【文献阅读】Perceptual Generative Adversarial Networks for Small Object Detection –CVPR-2017
- 目标检测“Perceptual Generative Adversarial Networks for Small Object Detection”
- 目标检测“Perceptual Generative Adversarial Networks for Small Object Detection”
- 目标检测“Perceptual Generative Adversarial Networks for Small Object Detection”
- Coupled Generative Adversarial Networks 阅读笔记
- Stacked Conditional Generative Adversarial Networks for Jointly Learning Shadow Detection and Shadow
- 论文阅读-《Relation Networks for Object Detection》
- GENERATIVE ADVERSARIAL NETWORKS FOR IMAGE STEGANOGRAPHY
- 【文献阅读】Densely Connected Convolutional Networks-best paper-CVPR-2017
- pytorch-generative-adversarial-networks
- 阅读小结:Unsupervised Representation with Deep Convolutional Generative Adversarial Networks
- BEGAN: Boundary Equilibrium Generative Adversarial Networks阅读笔记
- 论文阅读:Deep Neural Networks for Object Detection
- Feature Pyramid Networks for Object Detection 阅读笔记
- GANs实现(Generative Adversarial Networks)
- 阅读小结:Generative Adversarial Nets
- 论文阅读:Learning to Discover Cross-Domain Relations with Generative Adversarial Networks
- Visual Saliency Prediction with Generative Adversarial Networks
- hdu 2256 Problem of Precision(矩阵乘法+共轭公式)
- Nodejs笔记(2)——模块
- StoryBoard布局约束注意点
- error LNK2038: 检测到“_MSC_VER”的不匹配项问题
- 面试算法:波浪型数组的快速排序法
- 【文献阅读】Perceptual Generative Adversarial Networks for Small Object Detection –CVPR-2017
- ZOJ 1516 Uncle Tom's Inherited Land (二分图最大匹配)
- Andrew Ng机器学习课程笔记--week3
- Android之广播
- NET Core 使用 EF Code First
- bzoj4358 permu(莫队+线段树||莫队+并查集+分块)
- javascript
- 程序员,如何在工作之外,增加自己的收入?
- springcloud配置中心使用github作为文件存放地址