Roof type classification using one-shot learning approach

Recently, Convolutional Neural Network-based methods have been used frequently for roof-type classification on images taken from space. The most important problem with classification processes using these methods is that they require a large amount of training data. Usually, one or a few samples are enough for a human to recognise an object. Like the human brain, the One-Shot Learning approach aims to learn object categories with just one or a few training examples per class, rather than using huge amounts of data. In this study, roof-type classification was carried out with a few training examples using the one-time learning approach and the so-called Siamese Neural Network method. The images used for training were artificially produced due to the difficulty of finding roof data. Two different data sets consisting of real roof pictures were used for the test. The test and training data set consisted of three different roof types: Flat, Gable and Hip. Finally, the Siamese Neural Network model, which was trained with artificially produced pictures, achieved an average classification performance of 66% as a result of testing with real roof pictures. With the other data set prepared, a classification success of 85% was achieved. The same data were also tested with Convolutional Neural Networks and Support Vector Machines, and it was found that the highest success was achieved with the Siamese Neural Network model.