A multivariate information aggregation method for crowd density estimation and counting

Method	Part_A		Part_B
Method	MAE	MSE	MAE	MSE
MCNN^［16］	110.2	173.2	26.4	41.3
Switch-CNN^［17］	90.4	135.0	21.6	33.4
CSRNet^［19］	68.2	115.0	10.6	16.0
DUBNet^［26］	64.6	106.8	7.7	12.5
CAN^［21］	62.3	100	7.8	12.2
FF-CAM^［22］	71.0	109.8	10.3	15.8
HLMMNet^［23］	-	-	8.1	13.2
Ours	59.4	96.2	7.7	11.9

Table 2. Comparison of performance index results of multiple algorithms in UCF-QNRF
View table
View in Article
Table 2. Comparison of performance index results of multiple algorithms in UCF-QNRF
Method MAE MSE
MCNN^［16］ 277 426
Switch-CNN^［17］ 228 445
CAN^［21］ 107 183
DUBNet^［26］ 105.6 180.5
SFCN^［28］ 102.0 171.4
FF-CAM^［22］ 114.5 200.5
Ours 89.3 164.5

Table 3. Comparison of performance index results of multiple algorithms in the NWPU

View table

View in Article

Table 3. Comparison of performance index results of multiple algorithms in the NWPU

Method	Val		Test
Method	MAE	MSE	MAE	MSE
MCNN^［16］	218.5	700.6	232.5	714.6
CSRNet^［19］	104.8	433.4	121.3	387.8
CAN^［21］	93.5	489.9	106.3	386.5
SFCN^［28］	95.4	608.3	105.4	424.1
BL^［30］	93.6	470.3	105.4	454.2
HLMMNet^［23］	89.3	431.2	103.0	433.2
Ours	87.9	417.2	101.2	418.5

Table 4. Ablation results of MIE
View table
View in Article
Table 4. Ablation results of MIE
Method Val Test
MAE MSE MAE MSE
Ours（without MIE） 91.9 461.2 110.3 470.8
Ours （with MSS） 90.2 448.3 107.9 455.7
Ours （with SE） 90.7 450.2 109.7 459.4
Ours（with MSS and SE） 87.9 417.2 101.2 418.5

Table 5. Results of MCIA ablation experiment， model size and algorithm speed

View table

View in Article

Table 5. Results of MCIA ablation experiment， model size and algorithm speed

Method	Val		Test		Model size/MB	Average running speed of test image /s
Method	MAE	MSE	MAE	MSE	Model size/MB	Average running speed of test image /s
MCNN^［16］	218.5	700.6	232.5	714.6	19.2	2.8
CSRNet^［19］	104.8	433.4	121.3	387.8	16.2	2.1
MDPMIF^［23］	89.5	451.9	108.0	459.3	21.4	2.3
HLMMNet^［23］	89.3	431.2	103.0	433.2	28.7	2.6
Ours （空洞卷积）	87.5	416.9	100.7	418.3	27.3	2.5
Ours （步长卷积）	87.9	417.2	101.2	418.5	14.2	1.9
Ours （步长卷积单S-ASPP）	89.6	427.0	110.8	437.8	13.4	1.7

Tools

Get Citation

Copy Citation Text

Guanghui LIU, Qinmeng WANG, Xuanrun CHEN, Yuebo MENG. A multivariate information aggregation method for crowd density estimation and counting[J]. Optics and Precision Engineering, 2022, 30(10): 1228

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites

Paper Information

Category: Information Sciences

Received: Jan. 19, 2022

Accepted: --

Published Online: Jun. 1, 2022

The Author Email: LIU Guanghui (guanghuil@163.com)

DOI:10.37188/OPE.20223010.1228

Topics

laser devices and laser physics

Lasers and Laser Optics

Laser physics

laser manufacturing

Instrumentation, Measurement and Metrology

Table 1. Comparison of performance index results of multiple algorithm on ShanghaiTech

Table 1. Comparison of performance index results of multiple algorithm on ShanghaiTech

Table 2. Comparison of performance index results of multiple algorithms in UCF-QNRF

Table 2. Comparison of performance index results of multiple algorithms in UCF-QNRF

Table 3. Comparison of performance index results of multiple algorithms in the NWPU

Table 3. Comparison of performance index results of multiple algorithms in the NWPU

Table 4. Ablation results of MIE

Table 4. Ablation results of MIE

Table 5. Results of MCIA ablation experiment， model size and algorithm speed

Table 5. Results of MCIA ablation experiment， model size and algorithm speed