论文笔记：A survey of recent advances in CNN-based single image crowd counting and density estimation
2017-12-22 14:33
1086 查看
A survey of recent advances in CNN-based single image crowd counting and density estimation
目标
人群计数目标是计算一个密集场景下的人数，密度估计目标是将人群图像转换成密度图像，可以清晰地看出人群的分布。这两个问题已经被结合在一起。传统方法
基于检测的方法基于回归的方法
基于密度估计的方法
基于CNN的方法
对网络特性的分类：基础CNN
规模感知模型
上下文感知模型
多任务模型
对输入数据的分类：
基于块的方法
基于完整图像的方法
具体方法
Deep people counting in extremely dense crowds
基础CNN，基于块端到端的深度CNN回归模型，使用了AlexNet，将最后的全连接层的4096的神经元改成了一个单一神经元用于预测人数。
Fast crowd density estimation with convolutional neural networks
基础CNN，基于块将图像分为密度从高到低的五类，使用了Multi-stage ConvNet。使用了两个串联的分类器来提高效果。
Cross-scene crowd counting via deep convolutional neural networks
多任务模型，基于块提出学习图像到人数的映射，并在应用到新场景时使用符合场景的图像微调。训练时可选地训练以人数和密度估计为目标训练。
Learning to count with CNN boosting
基础CNN，基于块提出layered boosting和selective sampling方法。
End-to-end crowd counting via joint learning local and global count
上下文感知模型，基于完整图像直接接收整张图像作为输入，输出人数。组成部分：预训练的GoogLeNet，LSTM解码器和最后的全连接层。
Crowdnet: a deep convolutional network for dense crowd counting
上下文感知模型，基于块混合了深层和浅层全连接网络来预测密度图像，两种网络的组合可以有效减小视角偏差的影响。
Single-image crowd counting via multi-column convolutional neural network
规模感知模型，基于完整图像提出多列卷积模型MCNN，可用于任意密度和任意视角。同时使用了一种新的生成密度图像的方法，并公布了数据集ShanghaiTech。
Towards perspective-free object counting with deep learning
规模感知模型，基于完整图像提出HydraCNN，第一部分为多个CountingCNN，第二部分为全连接层，可以有效学习不同场景下的特征。
Switching convolutional neural network for crowd counting
规模感知模型，基于块在多列CNN的基础上加入一个基于VGG-16的选择器，选取最优的列。
Mixture of counting CNNs: adaptive integration of cnns specialized to specific appearance for crowd counting
规模感知模型，基于块使用多个expertCNN预测人数，和一个gatingCNN来给所有expertCNN设置权重。
Fully convolutional crowd counting on highly congested scenes
规模感知模型，基于完整图像仅使用一列CNN，预测时对不同规模的输出结果取平均值。
数据集
UCSDMall
UCF_CC_50
WorldExpo’10
ShanghaiTech
内容来自用户分享和网络整理，不保证内容的准确性，如有侵权内容，可联系管理员处理 
相关文章推荐
- [论文笔记] Money, glory and cheap talk: analyzing strategic behavior of contestants in simultaneous crowdsourcing contests on TopCoder.com (WWW, 2010)
- 人群密度估计--CNN-based Cascaded Multi-task Learning of High-level Prior and Density Estimation for Crowd
- 论文笔记：Single-Image Crowd Counting via Multi-Column Convolutional Neural Network
- 1607.CVPR-Joint Learning of Single-image and Cross-image Representations for Person ReID 论文笔记
- [论文笔记] Human computation: a survey and taxonomy of a growing field (CHI, 2011)
- 【论文笔记】物体检测系列 Light-Head R-CNN: In Defense of Two-Stage Object Detector
- A survey of task allocation and load balance in distributed system阅读笔记
- 论文笔记《A Survey of Model Compression and Acceleration for Deep Neural Networks》
- Fast single shot detection and pose estimation 论文笔记
- 论文笔记之：Let there be Color!: Joint End-to-end Learning of Global and Local Image Priors for Automatic
- Region-based Convolutional Networks for Accurate Object Detection and Segmentation----R-CNN论文笔记
- 【论文阅读笔记】Deep Learning based Recommender System: A Survey and New Perspectives
- 论文阅读笔记：Threat of Adversarial Attacks on Deep Learning in Computer Vision: A Survey
- [论文笔记] Leveraging the crowd as a source of innovation Does crowdsourcing represent a new model for product and service innovation? (SIGMIS-CPR, 2012)
- 论文笔记之：Let there be Color!: Joint End-to-end Learning of Global and Local Image Priors for Automatic
- 【论文学习笔记】Clustering by fast search and find of density peak
- 论文学习-系统评估卷积神经网络各项超参数设计的影响-Systematic evaluation of CNN advances on the ImageNet
- 论文笔记：An End-to-End Trainable Neural Network for Image-based Sequence Recognition and Its Application
- 论文阅读：Joint Learning of Single-image and Cross-image Representations for Person Re-identification
- 【论文阅读笔记】Deep Learning in Medical Imaging: Overview and Future Promise of an Exciting New Technique