A Survey on In-Domain and Cross-Domain Image Classification using SURF Features

Image classification algorithms using state-of-the art methods have grabbed much attention in computer vision area. In-domain classification assumes the testing data to be in the same domain as of the training data. Cross-Domain classification is a paradigm where testing data is from a different but related domain to the training data. We use Speeded-Up Robust Features (SURF) for feature extraction, which outperforms other feature detectors in its speed and robustness. This paper implements in-domain and cross-domain classification for standard image datasets using benchmark algorithms and compare the efficiency of the algorithms to choose the best one for a particular combination of source and target datasets. In this paper we try to implement in-domain and cross-domain performances for standard images datasets using benchmark algorithms. And compare the efficacy of the algorithms to choose the optimized one for a particular combination of datasets. We assume the dataset to be available upfront.