Large-Scale Coarse-to-Fine Object Retrieval Ontology and Deep Local Multitask Learning

Table 9

Top-k accuracy table between different deep architectures in category classification.

Model	Top-k accuracy
	1	2	3	4	5
FashionNet [8]	—	—	0.8258	—	0.9017
NASNet v3 [20]	0.6382	0.7739	0.8391	0.8817	0.9094
NASNet v3 APD	0.6384	0.7718	0.8388	0.8822	0.9123
ResNet-18 [19]	0.6549	0.7834	0.8433	0.8829	0.9078
ResNet-18 APD	0.6672	0.7942	0.8563	0.8922	0.9164
ResNet-101 [19]	0.6802	0.8027	0.8587	0.8912	0.9132
ResNet-101 APD	0.6895	0.8150	0.87188	0.9057	0.9275