Table 5 Feature extractor with Imagenet weights used in this work.

From: Multiple model visual feature embedding and selection method for an efficient ocular disease classification

Model

Input shape

Selected features size

Number of parameters

Memory (in bytes)

Feature layer

DenseNet201

(224, 224, 3)

1920

18,321,984

327,486,960

Global average pooling 2D

InceptionResNetV2

(229, 229, 3)

1536

54,336,736

379,140,364

Global average pooling 2D

EfficientNet B3

300, 300, 3

1536

10,783,535

361,891,419

Dropout