Table 1.

Overall results of FOANet and compared architectures on the Nvidia benchmark.

Method	Channels	Accuracy
FOANet	FOA + Sparse Fusion	91.28%
FOANet	FOA + Avg. Fusion	85.26%
Human	Colour	88.4%
Molchanov [23]	All (including infrared)	83.8%
Molchanov [23]	Depth + Flow	82.4%