Table 2 Performance comparison on the test set of the proposed model with different backbones in terms of BLEU, ROUGE, and CIDEr

From: A deep learning based automatic report generator for retinal optical coherence tomography images

Models

BLEU-1

BLEU-2

BLEU-3

BLEU-4

ROUGE

CIDEr

RETFound + LSTM

0.3597

0.2892

0.2441

0.2104

0.4274

1.4662

ResNet34

0.6073

0.5346

0.4807

0.4352

0.6094

3.0563

ResNet101

0.5858

0.5176

0.4656

0.4216

0.6080

3.1049

ResNet50 + LSTM

0.6125

0.5412

0.4853

0.4369

0.6220

3.2607

VGG19 + LSTM

0.5300

0.4626

0.4114

0.3677

0.5853

2.8696

Res2Net + LSTM

0.6083

0.5381

0.4835

0.4366

0.6255

3.3585

SeResNet50 + LSTM

0.6008

0.5321

0.4781

0.4315

0.6262

3.3524

DenseNet + LSTM

0.6089

0.5378

0.4825

0.4349

0.6229

3.2689

MORG(Proposed)

0.6099

0.5409

0.4871

0.4406

0.6310

3.4109

Back to article page