Survey on Scene Classification techniques

Abstract

Scene classification is emerging research areas in the field of computer vision, as it can be used in several applications such as surveillance, autonomous driving, robotics, and many more. Scene classification is to classify the scene as one of the categories predefined as a kitchen, coast, forest, living room, etc. This paper highlights the prevailing scene classification practices by summarizing the major categories of the scene classification available in the literature. Traditionally to classify the scene, researchers used local features such as Scale Invariant Feature Transform (SIFT), Speeded-Up Robust Features (SURF) or the global features like GIST from the image and classify using supervised learning method like Support Vector Machine (SVM). Nowadays deep learning based approaches are widely used as the approaches do not have to manually extract the features and can learn automatically. This paper discusses various traditional and deep learning based approaches for indoor and outdoor scene classification with their challenges. We also present analysis of state-of-the-art methods with their advantages and disadvantages for scene classification.