Crowd Counting by Adaptively Fusing Predictions from an Image Pyramid

Abstract

Because of the powerful learning capability of deep neural networks, counting performance via density map estimation has improved significantly during the past several years. However, it is still very challenging due to severe occlusion, large scale variations, and perspective distortion. Scale variations (from image to image) coupled with perspective distortion (within one image) result in huge scale changes of the object size. Earlier methods based on convolutional neural networks (CNN) typically did not handle this scale variation explicitly, until Hydra-CNN and MCNN. MCNN uses three columns, each with different filter sizes, to extract features at different scales. In this paper, in contrast to using filters of different sizes, we utilize an image pyramid to deal with scale variations. It is more effective and efficient to resize the input fed into the network, as compared to using larger filter sizes. Secondly, we adaptively fuse the predictions from different scales (using adaptively changing per-pixel weights), which makes our method adapt to scale changes within an image. The adaptive fusing is achieved by generating an across-scale attention map, which softly selects a suitable scale for each pixel, followed by a 1x1 convolution. Extensive experiments on three popular datasets show very compelling results.

arXiv commentcamera ready, accepted to BMVC 2018

References in corpus (2)

• Deformable Convolutional Networks
• Switching Convolutional Neural Network for Crowd Counting

Cited by in corpus (22)

• CNN-based Density Estimation and Crowd Counting: A Survey
• Neuron Linear Transformation: Modeling the Domain Shift for Crowd Counting
• Fine-Grained Crowd Counting
• Crowd Counting with Deep Structured Scale Integration Network
• ADCrowdNet: An Attention-injective Deformable Convolutional Network for Crowd Understanding
• Learning Spatial Awareness to Improve Crowd Counting
• Crowd Counting via Hierarchical Scale Recalibration Network
• Multi-Scale Attention Network for Crowd Counting
• Enhanced 3D convolutional networks for crowd counting
• PDANet: Pyramid Density-aware Attention Net for Accurate Crowd Counting
• AutoScale: Learning to Scale for Crowd Counting and Localization
• Context-Aware Crowd Counting
• Bayesian Multi-Scale Neural Network for Crowd Counting
• Content-aware Density Map for Crowd Counting and Density Estimation
• Improving the Learning of Multi-column Convolutional Neural Network for Crowd Counting
• Motion-guided Non-local Spatial-Temporal Network for Video Crowd Counting
• 3D Crowd Counting via Geometric Attention-guided Multi-View Fusion
• Multi-Level Attentive Convoluntional Neural Network for Crowd Counting
• Crowd Counting by Self-supervised Transfer Colorization Learning and Global Prior Classification
• AdaCrowd: Unlabeled Scene Adaptation for Crowd Counting
• Counting People by Estimating People Flows
• CCCNet: An Attention Based Deep Learning Framework for Categorized Crowd Counting