Step 5 · Classification
EfficientNet-B7
The largest EfficientNet, which scales depth, width and resolution together. The best performer.
ImageNet weights, last 30 layers unfrozen
From the paper
What we built
EfficientNet-B7 was selected as the final model because it outperformed the others. It integrated Conv2D layers, Global Average Pooling, weight-decay regularization and an adaptive learning-rate optimizer, with the last 30 layers unfrozen to refine its representations on this dataset.
- Transfer learning
- Compound scaling
- Weight decay
- Mixed precision
- Adaptive learning rate
- Eye photo in
- EfficientNet-B7 basecompound-scaled, ImageNet
- Conv 3×3, 512 filters → batch norm
- Global average pool
- Dense 1024 → dropout
- Dense 512 → layer norm
- Dense 256 → dropout
- Softmax, 5 classes
- EsotropiaExotropiaHypertropiaHypotropiaNormal
Published results
How EfficientNet-B7 did on the 81 test photos
- Accuracy
- 84.00%
- Precision
- 85.00%
- Recall
- 83.00%
- F1 score
- 84.40%
| Predicted class → | |||||
|---|---|---|---|---|---|
| True ↓ | Esotropia | Exotropia | Hypertropia | Hypotropia | Normal |
| Esotropia | 0.80 | 0.07 | 0.00 | 0.07 | 0.07 |
| Exotropia | 0.06 | 0.76 | 0.00 | 0.06 | 0.12 |
| Hypertropia | 0.06 | 0.00 | 0.94 | 0.00 | 0.00 |
| Hypotropia | 0.00 | 0.06 | 0.06 | 0.75 | 0.12 |
| Normal | 0.00 | 0.00 | 0.00 | 0.06 | 0.94 |
Each row is a true class; the numbers show where its test photos ended up. The diagonal is the share classified correctly.
It was strongest on hypertropia (94% correct) and weakest on hypotropia (75% correct).