Identification of Sheep Breeds by CNN- Based Pre-Trained Inceptionv3 Model
Murat Köklü, İlkay Çınar, Yavuz Selim Taşpınar, Ramazan Kursun
Abstract
It is very important for the farmers who produce sheep to know which sheep breeds are in the sheep herd and which breeds can provide more income than others in order to better manage their resources. For this purpose, we propose a CNN-based model that can detect the breed of sheep from facial images to detect sheep breeds quickly, effectively, and at a low cost. In this study, a dataset containing a total of 1680 images belonging to 4 different sheep breeds was used. The 2048 deep features of each of these images were extracted using the InceptionV3 CNN model and given as inputs to the kNN, SVM, and ANN classifiers. As a result of the classification processes, the highest accuracy in the classification of sheep breeds was obtained as 92.3% from the ANN model. When the results of the study are evaluated, it is possible to say that success has been achieved in the classification of sheep breeds with this study.