JRL-YOLO: A Novel Jump-Join Repetitious Learning Structure for Real-Time Dangerous Object Detection

Table 4

Comparison of MAP results tested on the VOC2007 dataset.

Precision	Classes	YOLOv2-Tiny	YOLOv3-Tiny	YOLOv4-Tiny	JRL-YOLO
AP (%)	Aeroplane	39.83	46.47	54.06	59.63
	Bicycle	53.10	54.24	62.13	59.90
	Bird	13.55	18.83	29.55	33.03
	Boat	15.83	18.31	26.73	31.49
	Bottle	3.82	8.79	16.40	18.56
	Bus	45.52	46.45	57.06	62.29
	Car	52.53	60.62	72.81	71.21
	Cat	35.10	35.77	50.12	53.93
	Chair	16.51	21.12	31.11	28.66
MAP (%)	Cow	29.16	31.62	47.76	42.56
	Dining table	27.37	28.42	35.47	44.62
	Dog	28.53	30.11	41.00	46.74
	Horse	55.40	57.84	65.57	67.99
	Motorbike	52.18	54.37	65.30	65.52
	Person	39.47	48.79	63.73	64.41
	Potted plant	6.84	10.87	18.77	21.31
	Sheep	31.36	40.85	46.70	42.97
	Sofa	21.15	25.72	36.54	44.79
	Train	46.75	51.56	59.32	68.24
	TV monitor	36.45	42.21	48.06	46.91
	—	35.43	39.59	46.50	48.88

The best result is shown in bold.