Artificial intelligence-based real-time identification of selected dairy cattle breeds

This thesis investigates the real-time detection of cattle breeds using artificial intelligence. A system was developed to accurately and rapidly classify different cattle breeds by leveraging image processing and deep learning techniques. The study addresses key challenges in the cattle farming industry and explores the potential contributions of AI and computer vision technologies to this sector. The dataset used comprises 1,208 images of various cattle breeds—Ayrshire, Brown Swiss, Jersey, Red Dane, and Holstein-Friesian—sourced from the Kaggle platform. These images were captured from diverse angles and under varying environmental conditions to ensure dataset diversity. Deep learning architectures such as ResNet50 and DenseNet121 were implemented, and their classification performance was evaluated using Python and the TensorFlow/Keras libraries. Among the models tested, the Convolutional Neural Network (CNN) achieved the highest accuracy in breed classification. The findings indicate that accurate breed identification is vital for optimizing feeding strategies and improving animal health monitoring. The proposed system has the potential to replace manual observation processes in large-scale farms, offering significant savings in time and labor. Thanks to its real-time capabilities, the system can provide immediate feedback based on the animals' condition, highlighting its potential to enhance operational efficiency in livestock management.