2019年,CV領域,值得一看的綜述性文章 !
本文僅用于學術交流,侵刪
問題:2019年,CV領域,你推薦哪些綜述性的文章?
來源:https://www.zhihu.com/question/355566860
知乎高質量回答
1、作者:Amusi
https://www.zhihu.com/question/355566860/answer/894352980
一直關注CV這一塊,下面分享幾個2019年比較好的CV綜述,方向涵蓋:目標檢測、圖像分割、目標跟蹤和超分辨率等
目標檢測
2019 四大目標檢測綜述論文:
Imbalance Problems in Object Detection: A Review
intro: under review at TPAMI
arXiv:?https://arxiv.org/abs/1909.00169
Recent Advances in Deep Learning for Object Detection
intro: From 2013 (OverFeat) to 2019 (DetNAS)
arXiv:?https://arxiv.org/abs/1908.03673
A Survey of Deep Learning-based Object Detection
intro:From Fast R-CNN to NAS-FPN
arXiv:https://arxiv.org/abs/1907.09408
Object Detection in 20 Years: A Survey
intro:This work has been submitted to the IEEE TPAMI for possible publication
arXiv:https://arxiv.org/abs/1905.05055
目標檢測更多論文詳見:
https://github.com/amusi/awesome-object-detection
圖像分割
Machine Learning Techniques for Biomedical Image Segmentation: An Overview of Technical Aspects and Introduction to State-of-Art Applications
arXiv :?https://arxiv.org/abs/1911.02521
Deep Semantic Segmentation of Natural and Medical Images: A Review
intro: 從 FCN(2014) 到 Auto-DeepLab(2019),本綜述共含179篇語義分割和醫(yī)學圖像分割參考文獻
arXiv:?https://arxiv.org/abs/1910.07655
Understanding Deep Learning Techniques for Image Segmentation
intro: 本綜述介紹了從2013年到2019年,主流的30多種分割算法(含語義/實例分割),50多種數(shù)據(jù)集,共計224篇參考文獻
arXiv :?https://arxiv.org/abs/1907.06119
目標跟蹤
A Review of Visual Trackers and Analysis of its Application to Mobile Robot
intro: 本目標跟蹤綜述共含185篇參考文獻!從傳統(tǒng)方法到最新的深度學習網(wǎng)絡
arXiv:?https://arxiv.org/abs/1910.09761
Deep Learning in Video Multi-Object Tracking: A Survey
intro: 38頁目標跟蹤綜述,含30多種主流算法,共計174篇參考文獻
arXiv:?https://arxiv.org/abs/1907.12740
超分辨率
A Deep Journey into Super-resolution: A survey
arXiv:?https://arxiv.org/abs/1904.07523
Deep Learning for Image Super-resolution: A Survey
arXiv:?https://arxiv.org/abs/1902.06068
2、作者:魏秀參
?https://www.zhihu.com/question/355566860/answer/896661195
自薦一篇“Deep Learning for Fine-Grained Image Analysis: A Survey“:
《超全深度學習細粒度圖像分析:項目、綜述、教程一網(wǎng)打盡》
鏈接:https://mp.weixin.qq.com/s/2pJt9hlUFhR6mo1ughKkiA
另,除文末提及的幾個具體future directions
Automatic Fine-Grained Models
Fine-Grained Few-Shot Learning
Fine-Grained Hashing
之外。實際上FGIA領域還有非常多新鮮好玩的問題和應用值得探索,如:
我們圍繞FGIA提出的一個目前最大的新零售場景商品數(shù)據(jù)集RPC:
https://zhuanlan.zhihu.com/p/55627416
在真實細粒度識別場景中不可避免的長尾分布問題:Long tailed problems
存在跨域差異(Domain adaptation)的細粒度圖像識別和檢索
……
期待更多CVer在FGIA領域作出有影響力的工作,更多FGIA信息可參見:
http://www.weixiushen.com/tutorials.html
-End-