RDMMFET: Representation of Dense Multimodality Fusion Encoder Based on Transformer

Table 2

Ablation studies on VQA v2.0 test-dev with iterations and layers of each encoder.

Module	Setting	Accuracy
Number of iterations	Epoch = 3	72.46
	Epoch = 4	72.59
	Epoch = 5	72.46
	Epoch = 6	72.22
Number of encoder layers		72.37
		72.35
		72.47
		72.44
		72.25
		72.37
		72.59