Table 4 The Accuracy and the corresponding standard (SD) of the pre-trained models are evaluated for categorizing OCT images related to AMD and DME.

From: Reinforcement-based leveraging transfer learning for multiclass optical coherence tomography images classification

(Pre-trained Model)

Dataset

1st Experiment

2nd Experiment

3rd Experiment

4rth Experiment

Median Accuracy

Standard Deviation

NASNetLarge

Training

0.8454

0.8470

0.8549

0.8586

0.8514

0.0036

Testing

0.8612

0.8830

0.8770

0.8734

0.8765

0.0056

MobileNetV2

Training

0.8817

0.8908

0.8837

0.8860

0.8855

0.0038

Testing

0.9080

0.9150

0.9110

0.9110

0.9112

0.0016

InceptionV3

Training

0.9287

0.9150

0.9280

0.9310

0.9256

0.0041

Testing

0.9480

0.9530

0.9518

0.9530

0.9514

0.0013

DenseNet201

Training

0.9278

0.9328

0.9340

0.9360

0.9326

0.0020

Testing

0.9490

0.9530

0.9610

0.9610

0.9560

0.0034

DenseNet121

Training

0.9130

0.9125

0.9012

0.9050

0.9079

0.0033

Testing

0.9330

0.9360

0.9350

0.9320

0.9340

0.0020

Inception- ResNetV2

Training

0.9379

0.9246

0.9343

0.9366

0.9333

0.0034

Testing

0.9600

0.9530

0.9564

0.9636

0.9582

0.0026

Back to article page