Using Deep Transfer Learning for Unsupervised Image Segmentation in Remote Sensing

Abstract

As multispectral image resolution has increased, generating accurate segmentation of these images can pose a significant challenge. Furthermore, creating accurate labeled data requires hours of manual segmentation. One solution to this problem is the application of deep learning algorithms which can learn non-linear trends in the data without significant preprocessing and be used for transfer learning. In this research, we demonstrate transfer learning on how a model trained on one dataset can be used to segment a different dataset. Using the U-Net deep learning algorithm, we first train our model on a dataset with class labels. We then use the trained model to extend a custom U-Net structure to transfer semantic knowledge from the previous training and adapt to the unknown images. Preliminary results indicate that there is a potential to achieve higher accuracy by using optimized loss functions suited for unsupervised learning along with pre-trained weights from the trained U-Net model.