USING COMPUTER VISION METHODS FOR PEOPLE DETECTION IN FOREST-STEPPE TERRAIN
A. L. Osipov, S. N. Tereshchenko
Novosibirsk State University of Economics and Management, Novosibirsk, Russia
Keywords: neural networks, artificial intelligence, unmanned aerial vehicle, forest area, human search, deep learning, augmentation, object detection
Abstract
The approach of applying machine learning methods to automatically detect people in forest-steppe areas in images is investigated. The artificial neural network technology is used to detect people from photographs. Deep learning methods of convolutional neural networks (YOLOv51, DenseNet, CenterNet, InceptionV3, Xception, and Faster R-CNN) are used in interaction with the "transfer learning" technique. Based on YOLOv51, a neural network is trained, which makes it possible to identify people using graphical images with an accuracy of 0.8795 on a test sample using the F1-score metric with a threshold value of 0.5.
|
