Learning Discriminative Mid-Level Patches for Fast Scene Classification

Abstract

Discriminative mid-level patch based approaches have become increasingly popular in the past few years. The reason of their popularity can be attributed to the fact that discriminative patches have the ability to accumulate low level features to form high level descriptors for objects and images. Unfortunately, state-of-the-art algorithms to discover those patches heavily rely on SVM related techniques, which consume a lot of computation resources in training. To overcome this shortage and apply discriminative part based techniques to more complicated computer vision problems with larger datasets, we proposed a fast, simple yet powerful way to mine part classifiers automatically with only class labels provided. Our experiments showed that our method, the Fast Exemplar Clustering, is 20 times faster than the commonly used SVM based methods while at the same time attaining competitive accuracy on scene classification.