Abstract:In a crowd density estimation dataset, the annotation of crowd locations is an extremely laborious task, and they are not taken into the evaluation metrics. In this paper, we aim to reduce the annotation cost of crowd datasets, and propose a crowd density estimation method based on weakly-supervised learning, in the absence of crowd position supervision information, which directly reduces the number of crowds by using the number of pedestrians in the image as the supervised information. For this purpose, we design a new training method, which exploits the correlation between global and local image features by incremental learning to train the network. Specifically, we design a parent-child network (PC-Net) focusing on the global and local image respectively, and propose a linear feature calibration structure to train the PC-Net simultaneously, and the child network learns feature transfer factors and feature bias weights, and uses the transfer factors and bias weights to linearly feature calibrate the features extracted from the Parent network, to improve the convergence of the network by using local features hidden in the crowd images. In addition, we use the pyramid vision transformer as the backbone of the PC-Net to extract crowd features at different levels, and design a global-local feature loss function ( ). We combine it with a crowd counting loss ( ) to enhance the sensitivity of the network to crowd features during the training process, which effectively improves the accuracy of crowd density estimation. The experimental results show that the PC-Net significantly reduces the gap between fully-supervised and weakly-supervised crowd density estimation, and outperforms the comparison methods on five datasets of ShanghaiTech Part A, ShanghaiTech Part B, UCF_CC_50, UCF_QNRF and JHU-CROWD++.

Towards using count-level weak supervision for crowd counting

Learning Crowd Scale and Distribution for Weakly Supervised Crowd Counting and Localization

Multi-Level Dynamic Graph Convolutional Networks for Weakly Supervised Crowd Counting

Region Feature Smoothness Assumption for Weakly Semi-Supervised Crowd Counting

Active Crowd Counting with Limited Supervision

A Weakly-Supervised Crowd Density Estimation Method Based on Two-Stage Linear Feature Calibration

Weakly Supervised Crowd-Wise Attention for Robust Crowd Counting

A Semi-supervised crowd counting method based on patch crowds statistics

Semi-Supervised Crowd Counting Via Multi-Task Pseudo-Label Self-Correction Strategy

Learning to Count via Unbalanced Optimal Transport

Semi-Supervised Crowd Counting Via Self-Training on Surrogate Tasks

Weakly-Supervised Scene-Specific Crowd Counting Using Real-Synthetic Hybrid Data

Hybrid Perturbation Strategy for Semi-Supervised Crowd Counting

Semi-supervised Crowd Counting Via Density Agency

Improving MLP-Based Weakly Supervised Crowd-Counting Network via Scale Reasoning and Ranking

Uniformity in Heterogeneity: Diving Deep into Count Interval Partition for Crowd Counting

Spatial Uncertainty-Aware Semi-Supervised Crowd Counting

Return of Small-Scale Crowd Counting via Fast and Accurate Semi-Supervised Least Squares Model

Glance to Count: Learning to Rank with Anchors for Weakly-supervised Crowd Counting

Hypergraph Association Weakly Supervised Crowd Counting

Densitytoken: Weakly-Supervised Crowd Counting with Density Classification