Introduction
Literature study
Transfer learning scheme for intelligent human counting system
Dataset collection
Dataset annotation
Dataset statistics
Human count | Training data | Validation data | Test data | Total |
---|---|---|---|---|
0 | 38 | 13 | 13 | 64 |
1 | 78 | 25 | 26 | 129 |
2 | 95 | 32 | 32 | 159 |
3 | 127 | 41 | 42 | 210 |
4 | 102 | 34 | 34 | 170 |
5 | 83 | 28 | 28 | 139 |
6 | 90 | 31 | 29 | 150 |
7 | 50 | 16 | 17 | 83 |
8 | 22 | 7 | 8 | 37 |
9 | 12 | 4 | 4 | 20 |
10 | 21 | 7 | 6 | 34 |
11 | 3 | 1 | 1 | 5 |
12 | 3 | 1 | 1 | 5 |
13 | 7 | 3 | 2 | 12 |
Total | 731 | 243 | 243 | 1217 |
Dataset | Number of images |
---|---|
UCF_CC_50 | 50 |
ShanghaiTech Part A | 482 |
ShanghaiTech Part B | 716 |
UCF_QNRF | 1535 |
WorldExpo’10 | 3980 |
RHC (our dataset) | 1217 |
Possible challenges
Experimental
Results and discussions
Quantitative analysis
Model | #Layers | Test MSE |
---|---|---|
AlexNet | 8 | 0.6240 |
VGG16 | 16 | 1.3762 |
GoogLeNet | 22 | 3.0069 |
ResNet18 | 18 | 2.2546 |
ResNet50 | 50 | 2.1044 |
ResNet101 | 101 | 2.0629 |
ResNet152 | 152 | 1.8185 |
DenseNet121 | 121 | 2.0378 |
Act. Cnt. | #Train data | AlexNet | VGG16 | Goog LeNet | ResNet | Dense Net121 | |||
---|---|---|---|---|---|---|---|---|---|
18 | 50 | 101 | 152 | ||||||
0 | 38 | 0.898 | 0.094 | 7.151 | 6.038 | 5.039 | 6.157 | 5.200 | 5.776 |
1 | 78 | 0.384 | 0.642 | 4.136 | 3.774 | 3.288 | 3.037 | 2.147 | 2.220 |
2 | 95 | 0.311 | 0.903 | 1.048 | 1.071 | 0.836 | 0.833 | 0.680 | 0.527 |
3 | 127 | 0.476 | 1.115 | 0.563 | 0.794 | 1.072 | 0.544 | 1.020 | 0.720 |
4 | 102 | 0.479 | 1.533 | 0.510 | 0.810 | 0.806 | 1.138 | 0.999 | 0.686 |
5 | 83 | 0.684 | 1.589 | 0.865 | 0.949 | 1.249 | 1.371 | 1.430 | 1.364 |
6 | 90 | 0.417 | 1.025 | 1.915 | 1.100 | 0.895 | 0.838 | 0.951 | 1.561 |
7 | 50 | 1.008 | 2.141 | 1.832 | 1.254 | 1.076 | 0.845 | 0.737 | 1.103 |
8 | 22 | 0.975 | 2.970 | 4.657 | 1.987 | 1.367 | 2.512 | 0.669 | 2.239 |
9 | 12 | 4.457 | 1.143 | 10.545 | 6.572 | 5.168 | 5.326 | 4.988 | 4.070 |
10 | 21 | 0.530 | 2.589 | 15.581 | 7.764 | 9.067 | 6.615 | 6.066 | 6.546 |
11 | 3 | 2.014 | 7.326 | 33.227 | 14.997 | 15.220 | 18.457 | 15.053 | 18.188 |
12 | 3 | 0.517 | 0.917 | 34.812 | 29.127 | 28.506 | 28.229 | 20.813 | 31.236 |
13 | 7 | 1.750 | 12.371 | 51.475 | 31.875 | 26.061 | 24.514 | 20.675 | 32.732 |