[caffe]深度学习之CNN检测object detection方法摘要介绍
2016-05-03 16:31
726 查看
[caffe]深度学习之CNN检测object detection方法摘要介绍
2015-08-17 17:44 3276人阅读 评论(1) 收藏 举报一两年cnn在检测这块的发展突飞猛进,下面详细review下整个cnn检测领域模型的发展,以及在时间性能上的发展。
一、RCNN
流程:
Extract region(off model) + extract features(on model) + classifyregions according feature (svm or softmax)性能:
精度:
二、SPP-NET
流程:
先做conv,再根据window提取特征。为什么rcnn不能也这么做呢?原因在于spp对不同尺度进行了max pool处理能更好的满足不同尺度window的特征表达。性能:
核心思想在全图只做一次conv,这个和overfeat的思想一致精度:
三、FAST-RCNN
流程:
引入了ROI层pooling,以及multi-task同时训练分类和检测框。性能:
Compared to SPPnet, Fast R-CNN trains VGG163× faster, tests 10× faster, and is more accurate.另外还额外提出了fc层SVD的思想
Vgg时间性能分析
精度:
The improvement of Fast R-CNN over SPPnetillustrates that even though Fast R-CNN uses single-scale training and testing,fine-tuning the conv layers provides a large improvement in mAP (from 63.1% to66.9%). Traditional R-CNN achieves a mAP of 66.0%. These resultsarepragmatically valuable given how much faster and easier Fast R-CNN is to trainand test, which we discuss next.
四、FASTER-RCNN
流程:
在fast-rcnn的基础上,借鉴了FCN的思路,将proposal阶段转化成一个layer加进了网络一起学习。性能:
cost-free for proposal精度:
our detection system has a frame rate of5fps (including all steps) on a GPU, while achieving state-of-the-art objectdetection accuracy on PASCAL VOC 2007 (73.2% mAP) and 2012 (70.4% mAP) using300 proposals per imagefrom: /article/3728876.html
相关文章推荐
- iOS使用Object-C自定义cordova插件(-)
- 2015 Objective-C 三大新特性
- 1.Objective-C简介
- JSONObject 和object之间的转换
- Python导入模块(包)的两种方式 TypeError: 'module' object is not callable
- 警告被视为错误 - 没有生成“object”文件
- NSArray中containsObject的用法
- Object与String的切换
- 通过 objc_setAssociatedObject alert 和 button关联 及传值
- a different object with the same identifier value was already associated with the session:
- html object元素
- select * from v$lock_object
- 【android】利用volley+Gson POST json请求来接收jsonObject并解析json数据
- Object-c------自定义构造方法
- Object-c------重写构造方法
- Programming with Objective-C(四)
- Object-c-----id
- Object-c------@property_and_@synthesize
- Object-c------NSString
- Objective和Swift,你该选择哪个