关闭

A-Fast-RCNN: Hard positive generation via adversary for object detection

354人阅读 评论(0) 收藏 举报

https://github.com/xiaolonw/adversarial-frcnn

  • Object detection requires the ability to be robust to illumination, deformation, occlusion and intra-class variations.

  • data-driven strategy – collect large-scale datasets which have object instances under different conditions.

  • The hope is that these examples capture all possible variations of a visual concept and the classifier can then effectively model invariance to them.

  • long tail : How can we sample such occlusions and deformations which lie on the tail?

https://mp.weixin.qq.com/s?__biz=MzA3Mjk0OTgyMg==&mid=2651123383&idx=1&sn=c2288947a721c5b88a5752bfac2ab5a2&chksm=84e6c7e6b3914ef0099c4cde6a1f2abad7623c1241a8f25cd2fbc18461b8aeb0fcb83778d0ce&mpshare=1&scene=1&srcid=0607ffLi3ZiOtO92OBb8BtLt&pass_ticket=P1YKbdSYRiC8HMaaqQDBffu2sMVg%2BMSXav4J1J6rknS%2FKu%2BERXnonUMhNLlNJYRB#rd

0
0
查看评论

目标检测“A-Fast-RCNN: Hard Positive Generation via Adversary for Object Detection”

如何训练一个目标检测器,对遮挡和形变鲁棒,目前的主要方法是增加不同情况的图像数据,但这些数据有时又特别少。作者提出使用对抗生成有遮挡或形变的样本,这些样本对检测器来说比较困难,使用这些困难的正样本训练可以增加检测器的鲁棒性。与Fast-RCNN比较,在VOC2007上,mAP增加了2.3%,VOC2...
  • cv_family_z
  • cv_family_z
  • 2017-04-13 11:41
  • 3164

对抗学习用于目标检测--A-Fast-RCNN: Hard Positive Generation via Adversary for Object Detection

A-Fast-RCNN: Hard Positive Generation via Adversary for Object Detection CVPR 2017 Caffe code : https://github.com/xiaolonw/adversarial-frcnn本文将对抗学习...
  • zhangjunhit
  • zhangjunhit
  • 2017-04-12 16:51
  • 5599

A-Fast-RCNN: Hard Positive Generation via Adversary for Object Detection

这篇论文作者在Fast RCNN的基础上,运用对抗生成网络GAN的思想,加入了两个对抗网络来加强Fast RCNN算法的鲁棒性,下面的内容是这篇文章的翻译,我已经修改过大部分的内容使得读起来比较通顺,接下来准备研究代码,根据作者的思想看看能否加入一点自己的东西。 代码:https://github....
  • qq_14839543
  • qq_14839543
  • 2017-05-16 18:21
  • 1098

阅读小结:A-Fast-RCNN: Hard Positive Generation via Adversary for Object Detection

arXiv: https://arxiv.org/pdf/1704.03414.pdf What: 1. 目标是去增强  检测器对于遮挡和形变 的泛化能力  2. 但是数据集中一般   遮挡和形变 的图像较少   3. 所以作者提出了 ...
  • Layumi1993
  • Layumi1993
  • 2017-06-16 20:35
  • 691

[论文笔记]A-Fast-RCNN: Hard Positive Generation via Adversary for Object Detection

[论文笔记]A-Fast-RCNN: Hard Positive Generation via Adversary for Object Detectionpaper一、论文思想训练一个目标检测器,对遮挡和形变鲁棒,目前的主要方法是增加不同场景下的图像数据,但这些数据有时又特别少。作者提出使用对抗生...
  • u011905800
  • u011905800
  • 2017-07-18 21:59
  • 255

【目标检测】Object Detection Fast RCNN 算法解析

Girshick, Ross. “Fast r-cnn.” Proceedings of the IEEE International Conference on Computer Vision. 2015. http://www.cv-foundation.org/openaccess/cont...
  • g110802008
  • g110802008
  • 2016-09-27 16:21
  • 1169

A-Fast-RCNN: Hard Positive Generation via Adversary for Object Detection 笔记

这篇paper主要是针对训练数据时,遮蔽和形变物体的数据集不一定很全面提出的,于是,想到可以用生成网络生成这种数据来进行训练。但是这种数据生成又是很困难的,作者就想到跳过生成数据,直接在Feature Map上做类似的生成。 和生成模型很类似,在Feature Map上引入Adversary去和D...
  • searobbers_duck
  • searobbers_duck
  • 2017-07-25 09:13
  • 261

对抗网络之目标检测应用:A-Fast-RCNN

对抗网络之目标检测应用:A-Fast-RCNN       论文:A-Fast-RCNN: Hard Positive Generation via Adversary for Object Detection 【点击下载】    ...
  • linolzhang
  • linolzhang
  • 2017-08-05 00:21
  • 2216

[论文阅读]R-FCN: Object Detection via Region-based Fully Convolutional Networks

arxiv上的一篇新论文,出自MSRA,目前还没有发表,今天刚读完,文章的缺点还要想一想,有空更新。原文链接:点击打开链接 本文是基于region based framework的一种新的detection方法,主要目的是通过移除最后的fc层进行加速。同时通过本篇论文,很好的将RCNN,f...
  • u012361214
  • u012361214
  • 2016-05-26 15:16
  • 12658

R-FCN: Object Detection via Region-based Fully Convolutional Networks

R-FCN: Object Detection via Region-based Fully Convolutional Networks背景介绍   R-CNN 系列的方法,如 SPPnet、Fast R-CNN、Faster R-CNN 等方法在 Object Detection 上取得了很大的...
  • u010807480
  • u010807480
  • 2016-07-10 22:29
  • 5975
    个人资料
    • 访问:4013次
    • 积分:271
    • 等级:
    • 排名:千里之外
    • 原创:23篇
    • 转载:0篇
    • 译文:1篇
    • 评论:0条