комп’ютерний зір

In this paper, we address the challenge of visual geo-localization from low-quality UAV imagery captured in real world environments. We propose a two-stage architecture, which includes Super-Resolution and visual geo-localization. We introduced novel, non-learnable Ensemble Super-Resolution (ESR) module, which first refines upscaled aerial frames, then seamlessly feeds the enhanced imagery into a visual geo-localization pipeline.

In this paper, we develop a new approach for detecting fire in images based on convolutional neural networks. Cascade structure, which provides improved efficiency of recognition in images with low resolution and objects that can visually resemble flames, was proposed. We have performed an experimental comparison with the modern method of objects detecting Faster R-CNN. As a result of the experiments, it was found that performance of fire recognition improved on average by 20%.