Table 1.

Results of food area detection.

Model name	Training mAPa (%)	Testing mAP (%)	Weight size (MB)	Training time (h)
YOLOv3b	81.1	74	236	178
YOLOv3-tiny	90	85	33.7	65
YOLOv3-tiny_3l	87.6	77	33.7	65
YOLOv4c	98.9	95	245	207
YOLOv4-tiny	70	74	23	64.9
YOLOv4-tiny_3l	85.1	75	23	64.3

^amAP: mean average precision. Input size was 608×608, and iteration numbers were 50,000.

^bYOLOv3: You Only Look Once, version 3.

^cYOLOv4 was selected from multiple models listed in the Table.