Multiple Context Learning Networks for Visual Question Answering

Table 6

Comparison with previous state-of-the-art methods on VQA v2.0 test dataset.

Model	Test-dev				Test-std
	All	Y/N	Num	Other	All
BUTD [13]	65.32	81.82	44.21	56.05	65.67
MFH [32]	68.76	85.31	49.56	59.89	-
Counter [33]	68.09	83.14	51.62	58.97	68.09
v-AGCN [17]	65.94	82.58	45.12	56.71	66.17
ReGAT [16]	70.27	86.08	54.42	60.33	70.58
DFAF [20]	70.22	86.09	53.32	60.49	70.34
MCAN [21]	70.63	86.82	53.26	60.72	70.90
MEDAN [22]	70.60	87.10	52.69	60.56	71.01
MCLN-LSTM	70.26	85.95	53.18	60.72	70.63
MCLN-BERT	71.05	87.43	53.28	61.08	71.48