A Generalized Loss Function for Crowd Counting and Localization
Authors:
- 1 June 2021
- conference paper
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1974-1983
Abstract
Previous work [40] shows that a better density map representation can improve the performance of crowd counting. In this paper, we investigate learning the density map representation through an unbalanced optimal transport problem, and propose a generalized loss function to learn density maps for crowd counting and localization. We prove that pixel-wise L2 loss and Bayesian loss [29] are special cases and suboptimal solutions to our proposed loss function. A perspective-guided transport cost function is further proposed to better handle the perspective transformation in crowd images. Since the predicted density will be pushed toward annotation positions, the density map prediction will be sparse and can naturally be used for localization. Finally, the proposed loss outperforms other losses on four large-scale datasets for counting, and achieves the best localization performance on NWPU-Crowd and UCF-QNRF.Keywords
This publication has 34 references indexed in Scilit:
- Spatiotemporal Modeling for Crowd Counting in Videos2017
- Generating High-Quality Crowd Density Maps Using Contextual Pyramid CNNs2017
- Switching Convolutional Neural Network for Crowd Counting2017
- Finding Tiny Faces2017
- SegNet: A Deep Convolutional Encoder-Decoder Architecture for Image SegmentationIEEE Transactions on Pattern Analysis and Machine Intelligence, 2017
- Deep Residual Learning for Image Recognition2016
- Single-Image Crowd Counting via Multi-Column Convolutional Neural Network2016
- Cross-scene crowd counting via deep convolutional neural networks2015
- Multi-source Multi-scale Counting in Extremely Dense Crowd Images2013
- Pedestrian Detection in Crowded Scenes2005