Table 11 Computational efficiency analysis of the DGOA-ensemble model on retinal fundus images dataset.
From: An enhanced diabetic retinopathy detection approach using optimized deep learning technique
Dataset size (number of images) | Method | Training time (s per epoch) | Inference time (ms per image) | Peak memory usage (GB) |
|---|---|---|---|---|
10,000 | Proposed DGOA + Ensemble | 42 | 6.8 | 7.5 |
Ref28. EfficientNetV2S | 55 | 8.2 | 9.1 | |
20,000 | Proposed DGOA + Ensemble | 85 | 7.3 | 13.2 |
Ref28. EfficientNetV2S | 110 | 9.0 | 15.8 | |
50,000 | Proposed DGOA + Ensemble | 215 | 8.1 | 28.4 |
Ref28. EfficientNetV2S | 280 | 10.5 | 33.7 |
