Table 11 Binary classification performance on APTOS 2019 datasets.
From: Knowledge distillation-based lightweight MobileNet model for diabetic retinopathy classification
Author and Year | Models | Trainable Parameters | Accuracy | Precision | Recall | F1-Score |
|---|---|---|---|---|---|---|
Chetoui et al., 202032 | EfficientNet-B7 | 66, 700, 000 | - | - | 98.1% | - |
BK Anoop et al., 202225 | Custom CNN | 184, 197, 154 | 94.6% | - | 86% | - |
Bala et al., 202228 | Custom CNN | 1, 100, 000 | 97.54% | 97.55% | - | 0.97 |
Nandakumar et al., 202231 | Modified DenseNet-121 | - | 96% | 93.51% | 98% | 0.98 |
Begriche et al., 202334 | fine-tuned XCeption | - | 99.8% | - | - | - |
ResNet152V2 + VIT (Teacher Model) | 145, 800, 000 | 95.15% | - | - | - | |
Islam et al., 202339 | XCeption + CBAM (Student Model) | 21, 400, 000 | 99% | - | - | - |
Tuncel et al., 202571 | VGG16 | - | 97% | 97% | 97% | 97% |
Naveen et al., 202572 | EffNet-SVM | - | 97% | 97% | 97% | 97% |
Teacher Model | MobileNet | 279, 378 | 99.45% | 99.45% | 99.45% | 99.45% |
Student without KD | Reduced parameter MobileNet | 71,362 | 94.73% | 94.73% | 94.73% | 94.73% |
Student with KD (Proposed Model) | Reduced parameter MobileNet | 71,362 | 98.36% | 98.36% | 98.36% | 98.36% |
