An Investigation of Deep Learning Object Recognition on Dangerous Dog Breeds

Abstract

Many deep learning approaches have been developed to solve artificial intelligence problems with deep learning architectures. Due to its powerful feature extraction and learning capabilities, it is frequently preferred in object recognition processes. Detection of dogs, which is one of the most preferred pets today, is important for different purposes. It is preferred in analyzes made on the basis of gender. In this article, deep learning methods and deep learning and segmentation methods are used together to detect the dog in a data set consisting of 3 different dangerous dog breeds. In the results obtained, it was seen that the accuracy rate increased to 88.33% with the tissue segmentation method used before NasNetLarge.