Table 6 Running times for the different approaches.
From: A Parallel Adaboost-Backpropagation Neural Network for Massive Image Dataset Classification
(a) Training time (s) | ||||
|---|---|---|---|---|
Classification approach | Image category | 1,000 images | 5,000 images | 15,000 images |
RF | 10 | 56 | 373 | 5,874 |
30 | 58 | 388 | 6,137 | |
100 | 67 | 404 | 6,581 | |
SVM | 10 | 55 | 371 | 5,869 |
30 | 58 | 386 | 6,130 | |
100 | 68 | 402 | 6,573 | |
Adaboost-BP | 10 | 56 | 372 | 5,872 |
30 | 59 | 388 | 6,135 | |
100 | 67 | 403 | 6,579 | |
The method of Shi et al. | 10 | 55 | 370 | 5,431 |
30 | 59 | 391 | 5,955 | |
100 | 68 | 401 | 6,417 | |
Parallel BP | 10 | 12 | 47 | 139 |
30 | 14 | 50 | 149 | |
100 | 19 | 54 | 155 | |
Parallel Adaboost-BP | 10 | 11 | 45 | 129 |
30 | 14 | 49 | 133 | |
100 | 18 | 52 | 148 | |
(b) Testing time (ms) | ||||
RF | 10 | 4 | 6 | 10 |
30 | 4 | 7 | 13 | |
100 | 6 | 11 | 16 | |
SVM | 10 | 4 | 6 | 9 |
30 | 4 | 7 | 12 | |
100 | 6 | 10 | 15 | |
Adaboost-BP | 10 | 4 | 6 | 9 |
30 | 4 | 7 | 12 | |
100 | 6 | 10 | 16 | |
The method of Shi et al. | 10 | 4 | 5 | 9 |
30 | 5 | 7 | 11 | |
100 | 5 | 10 | 14 | |
Parallel BP | 10 | 1 | 2 | 4 |
30 | 1 | 2 | 4 | |
100 | 1 | 3 | 4 | |
Parallel Adaboost-BP | 10 | 1 | 2 | 3 |
30 | 1 | 2 | 3 | |
100 | 1 | 2 | 4 | |
