Штучний інтелект

Анотація: The paper proposes a technique based on an ensemble of models for identifying objects of remote sensing of the Earth. As a model, it is proposed to use a multilayer convolutional neural network. It is proposed to perform the determination of the optimal hyperparameters of the model using the grid or random search methods using k-fold cross-validation. A method for improving the accuracy of identifying objects using an ensemble of neural networks is shown. The results of the experiment using remote sensing data of the Earth are presented. The task of identifying objects of two classes was solved, images obtained using a synthetic aperture radar were used as test data.

Ключові слова: convolutional neural network, neural network ensemble, remote sensing of the Earth, classification

Посилання: