Table 4.

The accuracy of multimodal gender classification model evaluated on the clean and noisy input data streams. ‘n/a’ stands for not available. The best accuracy results for the validation and test sets are boldfaced.

ID	Data streams			Accuracy (%)
	Visual	Audio	Thermal	Valid	Test
1	clean	clean	clean	89.6	96.0
2			noisy	87.9	95.6
3		noisy	clean	81.9	84.4
4			noisy	76.6	82.0
5	noisy	clean	clean	89.2	93.6
6			noisy	84.5	88.2
7		noisy	clean	67.5	65.8
8			noisy	55.2	50.1
9	clean	n/a	n/a	91.5	91.3
10	n/a	clean	n/a	90.1	94.8
11	n/a	n/a	clean	94.8	94.3