Table 17 Analysis of test data and features generated at the GAP and Output layers for clustering using Teacher and Student models (with and without knowledge distillation) for ternary classification on the Primary Dataset.
From: Knowledge distillation-based lightweight MobileNet model for diabetic retinopathy classification
Data/Layer | Model | Silhouette Score | Davies-Bouldin Index |
|---|---|---|---|
Test Data | - | 0.0097 | 8.39 |
GAP Layer | Teacher | 0.0901 | 2.38 |
Student (KD) | 0.0894 | 2.28 | |
Student (w/o KD) | 0.0763 | 2.81 | |
Output Layer | Teacher | 0.3447 | 0.9433 |
Student (KD) | 0.2683 | 1.0481 | |
Student (w/o KD) | 0.1962 | 1.1204 |
