Crowd Counting via Scale-Adaptive Convolutional Neural Network

Abstract : The task of crowd counting is to automatically estimate the pedestrian number in crowd images. To cope with the scale and perspective changes that commonly exist in crowd images, state-of-the-art approaches employ multi-column CNN architectures to regress density maps of crowd images. Multiple columns have different receptive fields corresponding to pedestrians (heads) of different scales. We instead propose a scale-adaptive CNN (SaCNN) architecture with a backbone of fixed small receptive fields. We extract feature maps from multiple layers and adapt them to have the same output size; we combine them to produce the final density map. The number of people is computed by integrating the density map. We also introduce a relative count loss along with the density map loss to improve the network generalization on crowd scenes with few pedestrians , where most representative approaches perform poorly on. We conduct extensive experiments on the ShanghaiTech, UCF_CC_50 and WorldExpo'10 datasets as well as a new dataset SmartCity that we collect for crowd scenes with few people. The results demonstrate significant improvements of SaCNN over the state-of-the-art.
Document type :
Conference papers
Complete list of metadatas

Cited literature [34 references]  Display  Hide  Download

https://hal.inria.fr/hal-01830946
Contributor : Miaojing Shi <>
Submitted on : Thursday, July 5, 2018 - 3:45:20 PM
Last modification on : Thursday, November 15, 2018 - 11:59:01 AM
Long-term archiving on : Monday, October 1, 2018 - 6:12:50 PM

File

wacv2018shi.pdf
Files produced by the author(s)

Identifiers

Citation

Lu Zhang, Miaojing Shi, Qiaobo Chen. Crowd Counting via Scale-Adaptive Convolutional Neural Network. WACV 2018 - IEEE Winter Conference on Applications of Computer Vision, Mar 2018, Lake Tahoe, United States. pp.1-9, ⟨10.1109/WACV.2018.00127⟩. ⟨hal-01830946⟩

Share

Metrics

Record views

653

Files downloads

336