CSRNet: Dilated Convolutional Neural Networks for Understanding the Highly Congested Scenes (2018-02-27T00:00:00.000000Z)

TL;DR

A network for Congested Scene Recognition called CSRNet is proposed to provide a data-driven and deep learning method that can understand highly congested scenes and perform accurate count estimation as well as present high-quality density maps.

Abstract

We propose a network for Congested Scene Recognition called CSRNet to provide a data-driven and deep learning method that can understand highly congested scenes and perform accurate count estimation as well as present high-quality density maps. The proposed CSRNet is composed of two major components: a convolutional neural network (CNN) as the front-end for 2D feature extraction and a dilated CNN for the back-end, which uses dilated kernels to deliver larger reception fields and to replace pooling operations. CSRNet is an easy-trained model because of its pure convolutional structure. We demonstrate CSRNet on four datasets (ShanghaiTech dataset, the UCF_CC_50 dataset, the WorldEXPO'10 dataset, and the UCSD dataset) and we deliver the state-of-the-art performance. In the ShanghaiTech Part_B dataset, CSRNet achieves 47.3% lower Mean Absolute Error (MAE) than the previous state-of-the-art method. We extend the targeted applications for counting other objects, such as the vehicle in TRANCOS dataset. Results show that CSRNet significantly improves the output quality with 15.4% lower MAE than the previous state-of-the-art approach.

Authors

Deming Chen

3 papers

Xiaofan Zhang

4 papers

Yuhong Li

3 papers

TL;DR

Abstract

Authors

References47 items

Machine learning on FPGAs to face the IoT revolution

High-performance video content recognition with long-term recurrent convolutional network for FPGA

Generating High-Quality Crowd Density Maps Using Contextual Pyramid CNNs

CNN-Based cascaded multi-task learning of high-level prior and density estimation for crowd counting

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

Switching Convolutional Neural Network for Crowd Counting

Rethinking Atrous Convolution for Semantic Image Segmentation

Object Region Mining with Adversarial Erasing: A Simple Classification to Semantic Segmentation Approach

Fully Convolutional Crowd Counting on Highly Congested Scenes

Towards Perspective-Free Object Counting with Deep Learning

Learning to Count with CNN Boosting

Xception: Deep Learning with Depthwise Separable Convolutions

End-to-end crowd counting via joint learning local and global count

CrowdNet: A Deep Convolutional Network for Dense Crowd Counting

Single-Image Crowd Counting via Multi-Column Convolutional Neural Network

YodaNN: An Ultra-Low Power Convolutional Neural Network Accelerator Based on Binary Weights

DeepLab: Semantic Image Segmentation with Deep Convolutional Nets, Atrous Convolution, and Fully Connected CRFs

Data-Driven Crowd Understanding: A Baseline for a Large-Scale Crowd Dataset

Shallow and Deep Convolutional Networks for Saliency Prediction

Going Deeper with Embedded FPGA Platform for Convolutional Neural Network

COUNT Forest: CO-Voting Uncertain Number of Targets Using Random Forest for Crowd Density Estimation

Multi-Scale Context Aggregation by Dilated Convolutions

STC: A Simple to Complex Framework for Weakly-Supervised Semantic Segmentation

Extremely Overlapping Vehicle Counting

Cross-scene crowd counting via deep convolutional neural networks

Learning Deconvolution Network for Semantic Segmentation

Crowded Scene Analysis: A Survey

Fully convolutional networks for semantic segmentation

Very Deep Convolutional Networks for Large-Scale Image Recognition

Caffe: Convolutional Architecture for Fast Feature Embedding

Multi-source Multi-scale Counting in Extremely Dense Crowd Images

Cumulative Attribute Space for Age and Crowd Density Estimation

ImageNet classification with deep convolutional neural networks

Learning to count with regression forest and structured labels

Pedestrian Detection: An Evaluation of the State of the Art

Learning To Count Objects in Images

Object Detection with Discriminatively Trained Part Based Models

Deconvolutional networks

Bayesian Poisson regression for crowd counting

Crowd analysis: a survey

Privacy preserving crowd monitoring: Counting people without people models or tracking

Histograms of oriented gradients for human detection

Image quality assessment: from error visibility to structural similarity

Robust Real-Time Face Detection

Object recognition from local scale-invariant features

Dropout: a simple way to prevent neural networks from overfitting

Crowd Counting and Profiling: Methodology and Evaluation

Field of Study

Journal Information

Name

Page