Table 4 The impact of different methods on detection precision.

Method	Backbone	AP	AP50	AP75	APS	APM	APL	Params (M)
DETR	ResNet50	0.576	0.913	0.648	0.436	0.564	0.597	41.56
Dab-detr	ResNet50	0.553	0.928	0.604	0.385	0.565	0.571	43.72
Dino	ResNet50	0.553	0.911	0.598	0.413	0.564	0.569	47.55
Dino	Swin-Large	0.590	0.941	0.653	0.497	0.590	0.607	218.23
Sparse r-cnn	ResNet101	0.542	0.907	0.581	0.409	0.553	0.559	124.94
Cascade r-cnn	ResNet50	0.545	0.889	0.597	0.358	0.553	0.562	69.16
YOLOx-L	CSPDarknet	0.573	0.903	0.631	0.409	0.566	0.591	54.21
YOLOv8-L	CSPDarknet	0.582	0.889	0.662	0.406	0.526	0.609	43.70
YOLOv10-L	CSPDarknet	0.588	0.896	0.659	0.401	0.558	0.612	24.42
DiffusionDet	ResNet101	0.541	0.898	0.582	0.353	0.559	0.559	62.12
RT-detr	ResNet50	0.595	0.918	0.656	0.450	0.576	0.623	41.97
Our model	Swin-Tiny	0.602	0.943	0.656	0.538	0.592	0.619	78.05

Quick links

Search