Cattle Breed Classification Using Deep Learning Process

Abstract

Cow fattening can be one of the programs to eliminate unemployment. which will meet the needs of the country, along with removing unemployment, and will provide an opportunity to export outside the country. This unemployment can also be eliminated by developing accurate cow identification application software for fattening using a Data Science and Machine Learning project. There is a wide opportunity to export beef abroad to meet the needs of the country. Massive development is possible in this sector by creating new entrepreneurs. Buying cattle for fattening and determining the breed of cattle is the most difficult task for a new entrepreneur. By photographing thousands of cows, the growth rate and profile will be saved and the future of the new cow will be known by providing accurate pictures so that an entrepreneur can select the right cow. Because choosing the right cow is the biggest challenge for a new breeder. Therefore, keeping in mind the new entrepreneur, we have come up with a data science and machine learning project which will determine which cow belongs to which class or breed and select the right cow at this speed. In this study, we used two algorithms MobileNet-V3 and DenseNet201. We collected 1186 images in our data set. We are dividing those data sets into 2 parts, train and test, and using 5 classes of cattle. The training data is 952 images, while the testing data is 234 images, and the image size is (224, 224). Using these two algorithms, MobileNetv3 and DenseNet201, we achieve 89.78% and 91.85% accuracy, respectively.