Table 11 Comparing architecture of base MobileNet and custom model30.
From: Software application in early blight detection in tomatoes using modified MobileNet architecture
Aspect | Custom mode | Base MobileNet |
|---|---|---|
Model Architecture | Custom architecture with MobileNet as base | MobileNet architecture |
Transfer learning | Uses MobileNet and Custom_Feature_Extraction_Blocks for feature extraction | MobileNet architecture |
Custom layers | Uses custom and dense layers | No custom layers |
Regularization | Applies L2 regularization to custom layers | Optional |
Data augmentation | Applies data augmentation to training data | No data augmentation |
Training | Trains custom model from scratch | Fine-tunes pre-trained MobileNet |
Training time | Longer training time because of custom layers and augmentation | Transfer learning results in faster training |
Interpretability | It is easier to interpret custom layers and visualize filters | Complex interpreting MobileNet layers |
Deployment | Heavier due to custom layers | Lightweight |
