Table 10 Detailed classification results of classification experiments integrating discriminative feature selection and random forest classification algorithms.

From: An efficient method for identifying surface damage in hydraulic concrete buildings

 

ResNet-18

MobileNet-v4

Precision

Recall

F1-score

Precision

Recall

F1-score

Crack

0.8215

0.8248

0.8231

0.8380

0.8550

0.8464

Fracture

0.8867

0.8510

0.8684

0.9285

0.8485

0.8867

Hole

0.9663

0.9329

0.9493

0.9469

0.9498

0.9484

Normal

0.8956

0.9603

0.9269

0.8853

0.9415

0.9125

 

EfficientNet-B0

RegNetY-800MF

Precision

Recall

F1-score

Precision

Recall

F1-score

Crack

0.8105

0.9289

0.8657

0.8872

0.8765

0.8818

Fracture

0.9477

0.8103

0.8737

0.9511

0.8782

0.9132

Hole

0.9789

0.9662

0.9725

0.9307

0.9699

0.9499

Normal

0.9390

0.9531

0.9460

0.9084

0.9525

0.9299

Back to article page