Scene Recognition using Multi-Label Multi-Resolution CNN

Abstract

Classification is one of the challenging issues of machine Learning and it is considered as one of the most complex issues in this area. In scene classification, feature extraction is one of the most important parts of every algorithm; because extraction of appropriate features will increase the accuracy of classification. Regarding the ambiguity and variety of the contents of images which have a poor appearance due to the changes of light and scale, scene recognition is more challenging than object recognition. In this paper, a new scene recognition method is proposed based on multi-label multi-resolution- CNN. Our method is evaluated by both indoor and outdoor scene datasets. The superior performance of our method is proved by simulation results and compared with other state of arts.