
论文笔记
lf14
学习者
展开
-
Deeply-Recursive Convolutional Network for Image Super-Resolution 笔记
首先介绍下,2016 CVPR的两篇SR oral是同一作者. Deeply-Recursive Convolutional Network for Image Super-Resolution(DRN) Accurate Image Super-Resolution Using Very Deep Convolutional Networks(VDSN)先介绍DRNAbstract: 如果训原创 2016-11-13 10:39:09 · 2394 阅读 · 0 评论 -
ONE-CLASS SVM FOR LEARNING IN IMAGE RETRIEVAL
One Class SVM 是指你的training data 只有一类positive (或者negative)的data, 而没有另外的一类。在这时,你需要learn的实际上你training data 的boundary。而这时不能使用 maximum margin 了,因为你没有两类的data。 所以呢,在这边文章中,“Estimating the support of a high-原创 2015-07-02 14:16:51 · 758 阅读 · 0 评论 -
Learning to Count Objects in Images
512x512LEARNING TO COUNT CELLS IN MICROSCOPY IMAGES----------Loading pre-trained SIFT codebook...----------Loading features precomputed at previous runMexifying MaxSubarray procedureNow原创 2016-02-21 16:20:31 · 1299 阅读 · 2 评论 -
《Vehicle Detection in High-Resolution Aerial Images via Sparse Representation and Superpixels》论文笔记
系统流程图:目标函数:里面用到的小trick:(1) 训练样本挑选:(2)计算效率改进:训练和测试旋转样本原创 2016-04-26 21:56:58 · 624 阅读 · 0 评论 -
《Vehicle Detection in High-Resolution Aerial Images Based on Fast Sparse Representation ...》论文笔记
系统流程图:这篇论文跟之前的一篇也有些差别,比如特征 = 低层特征(HOG+颜色名) + 高层特征(稀疏表示误差)=1388%% hog or sift size is set at 5x5 pixels LBP cell size is 7x7问题:样本挑选和high-order feature extraction 这里面的字典不一样,样原创 2016-05-01 18:33:48 · 740 阅读 · 0 评论 -
how hard can it be? Estimating the difficulty of visual search in an image
how hard can it be? Estimating the difficulty of visual search in an image摘要作者通过人对一幅画中是否有目标的响应时间定义图像的搜索难度。同时还分析图像属性,如图像尺寸、文件大小、边缘长度,超像素个数对视觉搜索难度产生的影响。文中采用卷积神经网络和回归模型对图像搜索难度进行预测,与ground-truth对比,取得较好效果。模原创 2016-07-12 22:42:28 · 805 阅读 · 0 评论 -
A Large Contextual Dataset for Classification, Detection and Counting of Cars with Deep Learning
这篇文章值得关注的点:1. This network combines residual learning with Inception-style layers and is used to count cars in one look. This is a new way to count objects rather than by localization or density estima原创 2016-10-13 21:00:22 · 611 阅读 · 0 评论 -
Fast and Accurate Image Upscaling with Super-Resolution Forests
这篇文章的大致想法是 训练随机森林,随机森林的叶子节点带有回归函数,所以他的目标函数,包括两个功能一个功能是把类似的图片分到相同的树节点,也就是11式中的第二项,类似聚类;另外一个功能叶节点的回归函数回归误差最小化.相比《Optimized Regressor Forest for Image Super-Resolution》,虽然两篇都使用到随机数森林,但是还是有差异,后者的系原创 2016-10-22 15:07:08 · 1383 阅读 · 2 评论 -
Fast and Accurate Image Upscaling with Super-Resolution Forests
组会报告的这篇,里面对树生成的过程还是有不理解的地方,待加强.原创 2016-10-27 13:26:09 · 1027 阅读 · 0 评论 -
Fast and Accurate Image Upscaling with Super-Resolution Forests
….原创 2017-04-17 00:04:02 · 606 阅读 · 1 评论