RGB-D Scene Classification using an Ensemble of Convolutional Neural Networks with Softmax Aggregation

Scene classification using a combination of depth images and color images has shown better performance in terms of classification accuracy. Convolutional neural networks are widely used for scene recognition tasks. The fusing of depth information with color information can be done with different convolutional neural networks. Ensemble methods are proven to increase classification accuracy. However, very few studies have been reported on the use of ensembles methods with convolutional neural networks. This study investigates a few architectures to combine depth modality with color modality. We also employ an ensemble method by aggregating outputs of the softmax activation function at the output layer. Our ensemble method provides better scene recognition accuracy compared to standalone architectures.