Table 9 Performance comparison of the fine-tuned EfficientNet-B0 model with and without stratified splitting.
Class / metric | With stratified splitting | Without stratified splitting | ||||
|---|---|---|---|---|---|---|
Precision | Recall | F1 Score | Precision | Recall | F1 Score | |
Apple Scab | 0.98 | 1.00 | 0.99 | 0.47 | 0.98 | 0.63 |
Black Rot | 1.00 | 1.00 | 1.00 | 0.94 | 1.00 | 0.97 |
Cedar Apple Rust | 1.00 | 0.96 | 0.98 | 1.00 | 0.48 | 0.65 |
Healthy | 1.00 | 1.00 | 1.00 | 0.99 | 0.70 | 0.82 |
Macro Avg F1 Score | 0.99 | 0.77 | ||||
Model-level Accuracy | 99.69% | 78.93% | ||||
