Table 2 Evaluation results of different detection models based on the constructed dataset, where AP50/AP75 denotes AP at IoU (Intersection over Union) = 0.5/0.75, APS/APM and ARS/ARL are AP for small/medium objects and AR for small/medium objects, respectively. AR1/AR10 represents AR given 1/10 detection per image.

From: A Lung Nodule Dataset with Histopathology-based Cancer Type Annotation

Model

AP

AP50

AP75

APS

APM

AR1

AR10

ARS

ARM

MobileNet50

0.4136

0.8479

0.3209

0.4087

0.8515

0.5194

0.5209

0.5182

0.8500

Faster R-CNN48

0.5429

0.9272

0.5357

0.5383

0.9010

0.6463

0.6567

0.6545

0.9000

SSD51

0.5574

0.8894

0.5547

0.5569

0.8515

0.6179

0.6254

0.6242

0.8500

RetinaNet52

0.5807

0.9144

0.5635

0.5774

0.9010

0.6657

0.6716

0.6697

0.9500

Yolo49

0.6424

0.9388

0.7694

0.6672

0.9505

0.7284

0.7507

0.7485

0.9500

Back to article page