92 - Autoencoders using transfer learning - Image colorization
The accuracy of deep learning depends on many factors including the amount of training data and training time (# of epochs). Unfortunately, for many microscopy (and other) applications we do not have the luxury of large amount of labeled training data. Certain tricks can be applied to augment training data but it still isn't enough. Transfer learning can help partially address this situation. Transfer learning allows for adapting networks and models trained for certain specific applications to other applications. For example, feature extraction portion of a network that's designed to discriminate cats and dogs can be used towards discriminating healthy vs malarial cells.
This tutorial explains the process of using transfer learning to design an autoencoder for image colorization.
The code from this video is available at: https://github.com/bnsreenu/python_fo...
