Opto-Electronic Engineering, Volume. 51, Issue 10, 240174(2024)
GLCrowd: a weakly supervised global-local attention model for congested crowd counting
Figures & Tables(12)
Fig. 3. Comparison of different methods. (a) Traditional convolution; (b) Standard self-attention; (c) Local attention
Table 1. Information of datasets
View tableView in ArticleTable 1. Information of datasets
Shanghai Tech UCF-QNRF UCF_CC_50 Part A Part B 图像数量 482 716 1535 50 平均尺寸 589×868 768×1024 2013×2902 2101×2888 平均人数 501 123.6 1279 1278 最大人数 3139 578 12865 4543 总人数 241677 88488 1251642 63974
Table 2. Information of experimental environment
View tableView in ArticleTable 2. Information of experimental environment
配置 参数 操作系统 Ubuntu 20.04.3 LTS (GNU/Linux 5.15.0-97-generic x86_64) 显卡型号 NVIDIA GeForce RTX 4090(×1) 显存大小 24 G
Table 3. Data of partial experimental parameters
View tableView in ArticleTable 3. Data of partial experimental parameters
实验参数 具体参数 权重衰减 5×10−4 训练总周期数 1200 优化器动量 0.95 学习率 1×10−5
Table 4. Experimental results on Shanghai Tech dataset
View tableView in ArticleTable 4. Experimental results on Shanghai Tech dataset
算法模型 Part A Part B MAE MSE MAE MSE HLNet[ 40 ]71.5 108.6 11.3 20 Transcrowd_gap[ 29 ]66.1 105.1 9.3 16.1 Transcrowd_token[ 29 ]69.0 116.5 10.6 19.7 MATT[ 41 ]80.1 129.4 11.7 17.5 OT_M[ 20 ]70.7 114.5 8.1 13.1 SUA_crowd[ 42 ]68.5 121.9 14.1 20.6 PDDNet[ 43 ]72.6 112.2 10.3 17.0 GLCrowd 64.887 104.411 8.958 16.202
Table 5. Experimental results on UCF_QNRF dataset
View tableView in ArticleTable 5. Experimental results on UCF_QNRF dataset
算法模型 MAE MSE HLNet[ 40 ]100.4 182.6 Transcrowd_token[ 29 ]98.9 176.1 Transcrowd_gap[ 29 ]97.2 168.5 OT_M[ 20 ]100.6 167.6 SUA_crowd[ 42 ]130.3 226.3 PDDNet[ 43 ]130.2 246.6 GLCrowd 95.523 173.453
Table 6. Experimental results on UCF_CC_50 dataset
View tableView in ArticleTable 6. Experimental results on UCF_CC_50 dataset
算法模型 MAE MSE MATT[ 41 ]355.0 550.0 CCTrans[ 31 ]245.0 343.0 SSGP_Crowd[ 44 ]355.0 505.0 SE Cycle GAN[ 45 ]373.4 528.8 CDPL_crowd[ 46 ]336.5 486.1 Transcrowd[ 29 ]272.2 395.3 CrowdFormer[ 47 ]218.8 330.4 PC-Net[ 48 ]217.3 309.7 GLCrowd 209.660 282.217
Table 7. Comparative results of ablation experiments
View tableView in ArticleTable 7. Comparative results of ablation experiments
局部 注意力 ConvFFN 回归 令牌 Part A Part B UCF_CC_50 MAE MSE MAE MSE MAE MSE 第一组 × × √ 75.260 125.441 10.003 19.315 240.120 327.796 第二组 √ × √ 67.403 108.298 9.428 18.446 236.233 305.122 第三组 √ √ √ 64.887 104.411 8.955 16.202 209.660 282.217
Tools
Get Citation
Copy Citation Text
Hongmin Zhang, Qianqian Tian, Dingding Yan, Lingyu Bu. GLCrowd: a weakly supervised global-local attention model for congested crowd counting[J]. Opto-Electronic Engineering, 2024, 51(10): 240174
Paper Information
Category: Article
Received: Jul. 24, 2024
Accepted: Oct. 8, 2024
Published Online: Jan. 2, 2025
The Author Email: Zhang Hongmin (张红民)
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20