Few Shot Scene Classification in Remote Sensing using Meta-Agnostic Machine
Dalal Alajaji, Haikel Alhichri
Abstract
Scene classification has become an important research topic in remote sensing (RS) field. The typical solution relies on labeling a large enough set of the RS scenes manually using expert opinion, then training the algorithm on this set to learn how to classify other new scenes correctly. The best performance deep learning models required a large labeled dataset for training. Accordingly, there is a great need to develop an intelligent machine learning algorithm that can learn to classify RS datasets containing new unseen classes from a few labeled samples only, which is known as few-shot machine learning. In this work, we develop a deep few-shot learning method for the classification of RS scenes. The proposed method is based on Model Agnostic Meta-Learning (MAML), one of the recently introduced and most popular meta-learning algorithms. In this paper, we report preliminary results using the UC Merced, OPTIMAL-31, and AID RS scene datasets.