Abstract:

Training neural network models is to learn them to solve specific problems from datasets. This process can take days or even weeks when a very large dataset is used, as is usually the case. Models must also be rebuilt from scratch as soon as the distribution of object spaces changes. One way to improve this process is to use transfer learning, which involves transferring knowledge learned from pre-trained models to new models with which there are similarities. This paper discusses the basic concepts, the steps to implement it, and its application to image classification.

Keywords: transfer learning, neural network, image classification