DETECTION OF DEER IN IMAGES BY COMPUTER VISION METHODS
S. N. Tereshchenko, A. L. Osipov
Novosibirsk State University of Economics and Management, Novosibirsk, Russia
Keywords: graphic images, reindeer herding, augmentation, neural networks, artificial intelligence, object detection, computer vision
Abstract
The approach of applying machine learning methods for automatic detection of deer individuals in images is studied. The neural network technology is used to accurately count the number of deer from photographs. Deep learning methods for convolutional neural networks (Resnet50, DenseNet, CenterNet, InceptionV3, Xception) are used in conjunction with the “transfer learning” technique. Based on the Faster R-CNN Resnet50 network, a neural network is trained, which makes it possible to determine deer individuals from graphic images with an accuracy of 0.91 on a sample using the F1-score metric with a threshold value of 0.6.
|