Compressed Wavelet Tensor Attention Capsule Network

Table 3

Classification accuracies (%) of six texture classification methods on three noisy texture datasets.

Method	Datasets
	CUReT	DTD	KTH-TIPS2-b
Wavelet CNNs	73.35 ± 1.35	56.18 ± 1.42	70.4 ± 1.65
T-CNN	67.64 ± 1.56	50.62 ± 1.44	68.8 ± 3.72
CapsNets	89.32 ± 1.08	67.08 ± 1.29	69.1 ± 1.48
FV-CNN	92.7 ± 1.54	71.5 ± 1.8	76.5 ± 3.52
SI-LCvMSP	89.6 ± 1.45	71.7 ± 1.38	91.58 ± 1.72
CWTACapsNet	99.1 ± 0.33	79.82 ± 0.54	96.9 ± 0.39