Scene Classification Algorithm Based on Semantic Segmented Objects

Scene classification is one of the important problems in image and video search and recommendation systems. We propose an efficient scene classification method for three different classes by detecting objects in the scene. For detecting objects in an image, pre-trained semantic segmentation model is used. A weight matrix which has bias values to determine a scene class statistically is constructed. Finally, we classify an image into one of three classes (i.e. indoor, nature, city) by using the designed weighting matrix. The proposed method achieved 92% of verification accuracy and improved over 2% when comparing to the existing convolutional neural network (CNN) models.