Robust Cattle Disease Classification Using Convolutional Neural Networks and Support Vector Machines

The existence of diseases in cattle offers many challenges in the practice of cattle rearing, not only in productivity but also in profitability. This paper examines the feasibility of employing a blend of CNN and SVM in persistently diagnosing cow diseases. This proposed model incorporates the ability of CNNs in feature extraction and SVMs in classification to boost the performance of illness diagnosis. The collection contains photos of affected cows with mastitis, foot and mouth disease, and bovine respiratory disease. These photographs were collected from veterinary clinics, farms, and internet sources. The preprocessing steps performed on the dataset were scaling and normalization with data augmentation techniques such as rotation, flipping, and contrast correction on the images. The developed hybrid model CNN-SVM reached the accuracy of classification 95%, which is significantly higher than compared to the classical ones (75%), or only CNN (88%) and only SVM (84%). Based on the results, the accuracy rate was 94%, the recall rate was 95%, and The F1 score was 94. 5%. The examination of the accuracy matrix led to the identification of high sensitivity and specificity of the model for all the identified illness groups. Further, the AUC-ROC values were steadily high all the time, indicating that the model has the best performance in classification between the classes. This paper has demonstrated that the CNN and SVM integrated approach is solid and efficient when classifying cattle diseases. This approach constitutes a stable and flexible instrument for diagnosing diseases in real time in agriculture.