Table 1.

The mean average precision (mAP) ± standard deviation (SD) and the detection (inference) speed/image ± standard deviation (SD) for YOLO and Faster R-CNN detectors across all proposed anchor boxes. The mean speed was calculated/averaged across 1,000 selected images.

Anchor boxes	YOLO		Faster R-CNN
	Mean average precision ± SD	Speed per image ± SD	Mean average precision ± SD	Speed per image ± SD
2	0.9695 ± 0.0269	$0.0112\pm 2.1\text{x}{10}^{-4}$	0.8987 ± 0.0713	$0.0419\pm 1.2\text{x}{10}^{-4}$
3	0.9888 ± 0.0094	$\mathbf{0.0128}\pm \mathbf{1.5}\text{x}{\mathbf{10}}^{-\mathbf{4}}$	0.9054 ± 0.0580	$0.0432\pm 6.3\text{x}{10}^{-4}$
4	0.9674 ± 0.0282	$0.0131\pm 2.8\text{x}{10}^{-4}$	0.9149 ± 0.0524	$\mathbf{0.0440}\pm \mathbf{2.9}\text{x}{\mathbf{10}}^{-\mathbf{4}}$
5	0.9663 ± 0.0288	$0.0144\pm 3.7\text{x}{10}^{-4}$	0.8919 ± 0.0842	$0.0445\pm 7.5\text{x}{10}^{-4}$
6	0.9657 ± 0.0295	$0.0158\pm 3.3\text{x}{10}^{-4}$	0.9039 ± 0.0766	$0.0476\pm 8.6\text{x}{10}^{-4}$