Lightweight YOLOv4 with Multiple Receptive Fields for Detection of Pulmonary Tuberculosis

Table 3

Evaluation results of each detection model on test set 2.

Model	Precision (%)	Recall (%)	mAP (%)	Miss detection rate (%)	Parameters (M)	Detection time (s)
U-net [6]	92.20	81.87	88.79	12	96.32	10.36
YOLOv3 [8]	89.14	76.08	74.92	22	61.53	9.63
Faster-RCNN [10]	94.96	83.66	92.40	8	136.65	11.62
YOLOv4 [12]	91.37	81.03	87.21	14	63.94	8.81
MIP-MY	96.59	85.50	95.32	6	33.91	5.72