Deep Learning for Computer Vision

Transfer learning or fine-tuning of a model

Transfer learning is the process of learning from a pre-trained model that was trained on a larger dataset. Training a model with random initialization often takes time and energy to get the result. Initializing the model with a pre-trained model gives faster convergence, saving time and energy. These models that are pre-trained are often trained with carefully chosen hyperparameters. 

Either the several layers of the pre-trained model can be used without any modification, or can be bit trained to adapt to the changes. In this section, we will learn how to fine-tune or transfer learning for a model that was trained on the ImageNet dataset with millions of classes.