Table 6.

Quantitative results on DFC15 multi-label dataset (%).

Model	m.$F_1$	m.$F_2$	m.${\mathrm{P}}_e$	m.${\mathrm{R}}_e$	m.${\mathrm{P}}_l$	m.${\mathrm{R}}_l$
VGGNet (Simonyan and Zisserman, 2014)	73.86	74.09	76.16	74.95	62.57	59.95
VGGNet-RBFNN (Zeggada et al., 2017)	72.21	73.02	74.08	74.42	60.82	66.58
CA-VGG-LSTM	75.46	75.85	77.95	76.95	73.56	59.19
CA-VGG-BiLSTM	76.25	76.93	78.27	78.30	74.99	64.31
GoogLeNet (Szegedy et al., 2015)	74.99	73.41	81.01	73.01	71.80	53.95
GoogLeNet-RBFNN (Zeggada et al., 2017)	73.38	72.62	78.46	72.94	64.62	63.22
CA-GoogLeNet-LSTM	75.67	75.46	79.08	76.12	70.22	60.65
CA-GoogLeNet-BiLSTM	78.25	76.80	83.97	76.52	82.98	61.04
ResNet-50 (He et al., 2016)	78.10	76.21	84.89	75.64	81.50	59.99
ResNet-RBFNN (Zeggada et al., 2017)	78.36	78.08	82.64	78.76	72.01	69.85
CA-ResNet-LSTM	78.78	76.65	85.66	75.84	83.83	60.05
CA-ResNet-BiLSTM	83.65	80.61	91.93	79.12	94.35	62.35