车辆计数--FCN-rLSTM: Deep Spatio-Temporal Neural Networks for Vehicle Counting in City Cameras

FCN-rLSTM: Deep Spatio-Temporal Neural Networks for Vehicle Counting in City Cameras
ICCV2017
https://arxiv.org/abs/1707.09476

利用监控相机来完成车辆计数，FCN+LSTM，使用 residual learning 将两者联系起来。FCN for pixel-level prediction and the strengths of LSTM for learning complex temporal dynamics

监控相机得到的视频具有以下几个特点：
1） Low frame rate 低帧率 1 fps to 0.3 fps
2）低分辨率 352 × 240, 320 × 240 or 704 × 480
3）高遮挡特别是交通拥堵时
4）Large perspective 导致车辆尺度范围大
以上四个特点导致用这些视频来进行车辆计数难度较大。

3 FCN-rLSTM for vehicle counting
这里我们采用 FCN 将 dense (pixel-level) feature 映射为车辆密度，避免了单个车辆的检测或跟踪。基于 FCN的密度估计可以让我们输入任意分辨率的图像，输出车辆密度图和输入图像尺寸一样大小。当前基于密度的计数方法直接对密度图求和得到总数。这么做误差比较大，因为 large perspective and oversized vehicles (big bus or big truck)，所以我们提出了 FCN-rLSTM network 来提高计数精度

3.1. FCN-rLSTM Model & Network Architecture

To accelerate training, FCN and LSTM are connected in a residual learning fashion as illustrated in Figure 4

3.2. Spatio-Temporal Multi-Task Learning
这里首先介绍了怎么生成训练样本的真值密度图，和人群计数的方法是一样的。 2D Gaussian kernels
接着定义了损失函数，包括两个部分

4 Experiments
Different configurations of FCN-rLSTM

Results comparison on WebCamT

Results comparison on TRANCOS dataset

Estimated density map for multiple cameras

Results comparison on UCSD dataset

车辆计数--FCN-rLSTM: Deep Spatio-Temporal Neural Networks for Vehicle Counting in City Cameras相关推荐

车辆计数“FCN-rLSTM: Deep Spatio-Temporal Neural Networks for Vehicle Counting in City Cameras”
使用深度时空神经网络,在低质量视频中计算车数目.为了使用视频时间信息,结合FCN和时间递归神经网络(LSTM),即FCN-rLSTM估计车辆密度和车辆计数.FCN进行像素级预测,LSTM学习复杂的时间 ...
【文本分类】Deep Pyramid Convolutional Neural Networks for Text Categorization
·阅读摘要: 本文提出了DPCNN(深度金字塔CNN)模型.在transformer.bert还没兴起的年代,模型越深效果越好,但是模型的复杂度会随着深度提升.粗略地说,DPCNN就是为了解决CN ...
PVANET: Deep but Lightweight Neural Networks for Real-time Object Detection
PVANET: Deep but Lightweight Neural Networks for Real-time Object Detection 传送门: 论文:https://www ...
经典DL论文研读(part2)--Deep Sparse Rectifier Neural Networks
学习笔记,仅供参考,有错必纠文章目录 Deep Sparse Rectifier Neural Networks Abstract Introduction Background Deep Rect ...
遥感图像-Deep Feature Alignment Neural Networks for Domain Adaptation of Hyperspectral Data高光谱数据深度特征对齐神经
Deep Feature Alignment Neural Networks for Domain Adaptation of Hyperspectral Data高光谱数据领域自适应的深度特征对齐神 ...
《Multi-scale Convolutional Neural Networks for Crowd Counting》论文笔记
Multi-scale Convolutional Neural Networks for Crowd Counting 论文地址论文翻译 ABSTRACT 1. INTRODUCTION 2. M ...
快速人群密度估计--Multi-scale Convolutional Neural Networks for Crowd Counting
Multi-scale Convolutional Neural Networks for Crowd Counting https://arxiv.org/abs/1702.02359 对于人群密度 ...
[C1W3] Neural Networks and Deep Learning - Shallow neural networks
第三周:浅层神经网络(Shallow neural networks) 神经网络概述(Neural Network Overview) 神经网络的表示(Neural Network Represent ...
【文献学习】Analysis of Deep Complex-Valued Convolutional Neural Networks for MRI Reconstruction
目录 1 简介和创新点 2 模型 2.1 复数激活函数 2.2 系统模型 3 模型参数 4 实验分析 4.1 激活函数分析 4.2 网络宽度分析 4.3 网络深度分析 4.4 网络结构分析 4.5 实 ...

车辆计数--FCN-rLSTM: Deep Spatio-Temporal Neural Networks for Vehicle Counting in City Cameras

车辆计数--FCN-rLSTM: Deep Spatio-Temporal Neural Networks for Vehicle Counting in City Cameras相关推荐

最新文章

热门文章