DeepSSN: A deep convolutional neural network to assess spatial scene similarity

Spatial-query-by-sketch is an intuitive tool to explore human spatial knowledge about geographic environments and support communication with scene database queries. However, traditional sketch-based search methods perform inadequately due their inability find hidden multiscale map features from mental sketches. In this research, we propose a deep convolutional neural network, namely the Deep Spatial Scene Network (DeepSSN), better assess similarity. DeepSSN, triplet loss function designed as comprehensive distance metric similarity assessment. A positive negative example mining strategy ensure consistently increasing distinction of triplets during training process. Moreover, develop prototype system using proposed in which users input queries via sketch maps can automatically augment data. The model validated multisource conflated data including 131,300 labeled samples after augmentation. empirical results demonstrate that DeepSSN outperforms baseline k-nearest neighbors, multilayer perceptron, AlexNet, DenseNet, ResNet mean reciprocal rank precision metrics. This research advances information retrieval studies by introducing novel learning method tailored