Transfer-Learning

Explored two different types of Transfer Learning.

 The first approach is using a pre-trained CNN as a fixed feature extractor.

In this technique, all layers of the CNN are frozen except for the last fully-connected layer. This last layer is changed to suit the task at hand.Took a pretrained model and replace the last fully connected layer which classifies images into 1000 classes into a new classifier that is adapted to classify images into 5 classes.

 In the second approach, once again, replaced the final fully connected layer of the network with a 5-class classifier.

However, this time trained the last layer for CIFAR-10 dataset.

 In both cases, Took a small 8,000 image dataset comprising 5 classes

and trained the CNN on this small dataset.

 Use Resnet as pre-trained model.

#Keras #Cifar-10 #CNN