The utilization of neural networks in dog emotion classification has great potential to improve the understanding of pet emotions. The goal is to develop a dog emotion classification system. This is important due to the lack of public ability to recognize and understand dog emotions. Neural networks able to create learning models can be used for decision-making, thus helping to reduce the risk of dangerous dog attacks. CNN itself is part of neural networks, where the CNN model has a higher accuracy rate of 74.75% compared to ResNet 65.10% and VGG 68.67%. Modeling using ROC-AUC shows the model's ability to distinguish emotion classes well. Angry has the highest AUC of 0.97, happy 0.93 and sad 0.96. While relaxed has the lowest AUC of 0.92. Classification report results show model has the highest precision and F1-Score values in angry class, while the highest recall value is in sad class.
Copyrights © 2024