CSRNet: Dilated Convolutional Neural Networks for Understanding the Highly Congested Scenes

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.

Accepted by CVPR 2018

References in corpus (8)

• Very Deep Convolutional Networks for Large-Scale Image Recognition
• Rethinking Atrous Convolution for Semantic Image Segmentation
• Caffe: Convolutional Architecture for Fast Feature Embedding
• Learning Deconvolution Network for Semantic Segmentation
• Shallow and Deep Convolutional Networks for Saliency Prediction
• CrowdNet: A Deep Convolutional Network for Dense Crowd Counting
• Switching Convolutional Neural Network for Crowd Counting
• FCN-rLSTM: Deep Spatio-Temporal Neural Networks for Vehicle Counting in City Cameras

Cited by in corpus (12)

• From Open Set to Closed Set: Counting Objects by Spatial Divide-and-Conquer
• Stacked Pooling: Improving Crowd Counting by Boosting Scale Invariance
• Enhanced 3D convolutional networks for crowd counting
• Learning Short-Cut Connections for Object Counting
• Multi-Resolution Fusion and Multi-scale Input Priors Based Crowd Counting
• LDC-Net: A Unified Framework for Localization, Detection and Counting in Dense Crowds
• SOFA-Net: Second-Order and First-order Attention Network for Crowd Counting
• PSC-Net: Learning Part Spatial Co-occurrence for Occluded Pedestrian Detection
• A Light-Weighted Convolutional Neural Network for Bitemporal SAR Image Change Detection
• Efficient Crowd Counting via Structured Knowledge Transfer
• Wisdom of (Binned) Crowds: A Bayesian Stratification Paradigm for Crowd Counting
• Recurrent Distillation based Crowd Counting