Face Detection Method based on Lightweight Network and Weak Semantic Segmentation Attention Mechanism

Table 5

Comprehensive comparison between this method and other mainstream methods.

Method	mAP/%	FPS	Model Size/MB
RetinaNet50	85.2	3.1	296.8
Cascade R-CNN	94.9	1.6	553.6
YOLOv3	93.4	3.3	255.8
ATSS	92.9	3.3	257.1
PAA	93.2	2.1	253.6
YOLOv4	96.7	3.6	242.9
YOLOv4-tiny	89	16.2	23.7
Proposed	96.9	12.0	8.4