Table 1: Slice-by-slice performance comparison between the baseline and proposed methods
Methods
mAP (%)
Precision (%)
Recall (%)
@10
@50
@10
@50
@10
@50
ResNet50(Baseline)
37.9
29.7
32.2
24.6
42.8
34.6
FocalNet
39.7
30.7
35.0
26.7
43.2
34.4
ResNet50+DRC
49.9
38.3
42.5
35.1
50.2
41.6
FocalNet+DRC
52.8
44.3
50.5
41.9
55.4
47.0