Full Model Architecture
Understand the full ResNet model architecture, focusing on how batch normalization provides regularization and why the number of filters increases at block layers. Learn how global average pooling and fully connected layers produce logits efficiently for large datasets. This lesson helps you grasp ResNet's design principles for building powerful image classifiers.
We'll cover the following...
We'll cover the following...
Chapter Goals:
- Learn about the full model architecture of ResNet
A. Regularization
For extremely deep models like ResNet, it is vital that we regularize the model, i.e. apply techniques to prevent overfitting. In the CNN and SqueezeNet sections, we used dropout for regularization. However, in ResNet we don't use dropout because it is normally not necessary to use dropout together with batch normalization.
The creators of batch ...