Research Article
DeepLumina: A Method Based on Deep Features and Luminance Information for Color Texture Classification
Table 8
Comparison with CNN-FC models with and without luminance for DTD and FMD Datasets.
| | Dataset | DTD | FMD |
| | Pretrained | RGB | RGB + Y | RGB | RGB + Y | | CNN-FC models | RGB | Luminance | Luminance |
| | MobileNet | 59.49 | 65.50 | 68.50 | 85.00 | | ResNet50 | 63.20 | 72.12 | 75.21 | 88.00 | | ResNet101 | 64.10 | 70.50 | 75.56 | 87.50 | | DenseNet201 | 60.50 | 68.50 | 70.50 | 80.25 | | AlexNet | 61.50 | 67.11 | 65.20 | 81.25 | | VGG19 | 64.20 | 72.61 | 77.50 | 84.00 | | Inceptionv3 | 62.50 | 69.50 | 78.50 | 88.50 |
|
|
The best accuracy obtained for CNN-FC models with and without luminance for DTD and FMD datasets is indicated in bold.
|
