Texture Synthesizability Assessment via Deep Siamese-Type Network

Table 1

Performance comparison between our network and other architectures. The dataset is the ETH Synthesisability dataset [3].

Approach	One patch	Multipatches	PLCC	SROCC	RMSE
Single column shallow	Y		0.112	0.087	0.921
Single column VGG-16	Y		0.413	0.305	0.507
Siamese VGG-16	Y		0.532	0.401	0.421
Siamese VGG-16		Y	0.614	0.522	0.335
Ours		Y	0.785	0.723	0.250

Bold entries are the best performers.