Table 4.

Comparison of detection speed of different models

Models	Inference speed	Parameters	FLOPs
Mask R-CNN	15.0 s	43.8 M	499.8G
Swin-S Mask R-CNN	9.81 s	69.1 M	723.5G
YOLO V4	0.419 s	63.9 M	362.1G
YOLOV5-L	0.413 s	46.5 M	292.5G
YOLOX-L	0.454 s	54.2 M	397.8G
YOLO POD	0.462 s	54.2 M	394.9G

All the models are tested at 1024 × 1024 resolution, and batch = 1 on 2080Ti