Histograms of Sparse Codes for Object Detection用于目标检测的稀疏码直方图
Abstract
Object detection has seen huge progress in recent years, much thanks to the heavily-engineered Histograms of Oriented Gradients (HOG) features. Can we go beyond gradients and do better than HOG? We provide an affirmative answer by proposing and investigating a sparse representation for object detection, Histograms of Sparse Codes (HSC).We compute sparse codes with dictionaries learned from data using K-SVD, and aggregate per-pixel sparse codes to form local histograms. We intentionally keep true to the sliding window framework (with mixtures and parts) and only change the underlying features. To keep training (and testing) efficient, we apply dimension reduction by computing SVD on learned models, and adopt supervised training where latent positions of roots and parts are given externally e.g. from a HOG-based detector. By learning and using local representations that are much more expressive than gradients, we demonstrate large improvements over the state of the art on the PASCAL benchmark for both rootonly and part-based models.
Histograms of Sparse Codes for Object Detection用于目标检测的稀疏码直方图的更多相关文章
- CVPR2020论文解读:3D Object Detection三维目标检测
CVPR2020论文解读:3D Object Detection三维目标检测 PV-RCNN:Point-Voxel Feature Se tAbstraction for 3D Object Det ...
- Mask R-CNN用于目标检测和分割代码实现
Mask R-CNN用于目标检测和分割代码实现 Mask R-CNN for object detection and instance segmentation on Keras and Tenso ...
- 带你读AI论文丨用于目标检测的高斯检测框与ProbIoU
摘要:本文解读了<Gaussian Bounding Boxes and Probabilistic Intersection-over-Union for Object Detection&g ...
- Sparse R-CNN: End-to-End Object Detection with Learnable Proposals 论文解读
前言 事实上,Sparse R-CNN 很多地方是借鉴了去年 Facebook 发布的 DETR,当时应该也算是惊艳众人.其有两点: 无需 nms 进行端到端的目标检测 将 NLP 中的 Transf ...
- Towards Universal Object Detection by Domain Attention
论文及代码 论文地址:https://arxiv.org/abs/1904.04402 代码:http://www.svcl.ucsd.edu/projects/universal-detection ...
- zz——Recent Advances on Object Detection in MSRA
本文由DataFun社区根据微软亚洲研究院视觉组Lead Researcher Jifeng Dai老师在2018 AI先行者大会中分享的<Recent Advances on Object D ...
- [论文理解] Acquisition of Localization Confidence for Accurate Object Detection
Acquisition of Localization Confidence for Accurate Object Detection Intro 目标检测领域的问题有很多,本文的作者捕捉到了这样一 ...
- ICCV2019论文点评:3D Object Detect疏密度点云三维目标检测
ICCV2019论文点评:3D Object Detect疏密度点云三维目标检测 STD: Sparse-to-Dense 3D Object Detector for Point Cloud 论文链 ...
- Adversarial Examples for Semantic Segmentation and Object Detection 阅读笔记
Adversarial Examples for Semantic Segmentation and Object Detection (语义分割和目标检测中的对抗样本) 作者:Cihang Xie, ...
随机推荐
- Spoj 2798 Qtree3
一棵结点为黑色或白色的树,初始都是白色的.有两种操作 1 将一个结点换颜色 2 询问从根到结点u路径上面的第一个黑色点,没有则输出-1 InputIn the first line there are ...
- 【HBase】四、HBase的安装及命令行接口
通过前面的介绍,对HBase数据模型,运行机制等理论基本了解,接下来从实践的角度介绍HBase的安装以及其各种接口的使用方法. 1.HBase的安装 HBase安装很简单,和所有的Hadoop ...
- Mybatis-学习笔记(3)mapper配置文件
1.mapper配置文件常用的元素 parameterMap已经废弃,老式风格的参数映射. 2.select元素 映射查询语句.#{...}用于预处理语句参数,通过JDBC,这样一个参数在SQL中会由 ...
- 极*Java速成教程 - (6)
Java高级特性 String String是Java中的字符串类型,字符串类型在内存中是一个不可变的对象.如果要对字符串对象进行修改,如果是较少的修改可以使用+运算符,Java会自动进行优化,但如果 ...
- [BZOJ 3456]城市规划(cdq分治+FFT)
[BZOJ 3456]城市规划(cdq分治+FFT) 题面 求有标号n个点无向连通图数目. 分析 设\(f(i)\)表示\(i\)个点组成的无向连通图数量,\(g(i)\)表示\(i\)个点的图的数量 ...
- 【系统】win10锁屏后,护眼绿自动恢复解决
针对自己电脑(其他人的不晓得),win10锁屏后,重新登录,护眼绿会自动恢复成白色,查询资料需要修改注册表两个地方: 1.计算机\HKEY_CURRENT_USER\Control Panel\Col ...
- thinkphp整合Ueditor编辑器
编辑器下载地址:http://ueditor.baidu.com/website/download.html#ueditor 放在项目Public或者入口同级目录均可. 前台代码 <div cl ...
- 线程中断:Thread类中interrupt()、interrupted()和 isInterrupted()方法详解
首先看看官方说明: interrupt()方法 其作用是中断此线程(此线程不一定是当前线程,而是指调用该方法的Thread实例所代表的线程),但实际上只是给线程设置一个中断标志,线程仍会继续运行. i ...
- 自定义UICollectionViewLayout
UICollectionView在iOS6中第一次被介绍,也是UIKit视图类中的一颗新星.它和UITableView共享API设计,但也在UITableView上做了一些扩展.UICollectio ...
- nmblookup - 基于TCP/IP上的NetBIOS客户用于查询NetBIOS名字的程序
总览 SYNOPSIS nmblookup [-M] [-R] [-S] [-r] [-A] [-h] [-B <broadcast address>] [-U <unicast a ...