RDMMFET: Representation of Dense Multimodality Fusion Encoder Based on Transformer

Table 3

Comparison with the latest models on the VQA v2.0 data set.

Label	Method	Test-dev	Test-std
No pretraining	DFAF [8]	70.22	70.34
	MCAN [9]	70.63	70.90
	MUAN [38]	70.82	71.10
Pretraining	ViLBERT [23]	70.55	70.92
	VisualBert [27]	70.80	71.00
	VL-BERT(base) [28]	71.16	-
	VL-BERT(large) [28]	71.79	72.22
	LXMERT [24]	72.42	72.54
	RDMMFET (ours)	72.59	72.67