DeepLumina: A Method Based on Deep Features and Luminance Information for Color Texture Classification

Table 10

Comparative Analysis for the FMD dataset.

Authors	Method	Accuracy
Song et al. [25]	FC-CNN	78.1 ± 1.6
Song et al. [25]	FV-CNN	80.2 ± 1.8
Song et al. [25]	FC-CNN + FV-CNN	83.2 ± 1.6
Simon et al. [18]	Deep features + SVM	84.50
Cimpoi et al. [27]	IFV + DeCAF	65.5
Bell et al. [58]	SIFT-IFV + fc7	69.6 ± 0.3
DeepLumina (proposed)	Deep Features (ResNet101) + luminance + SVM	90.15 ± 1.2

The best accuracy value is indicated in bold.