Presented at the 2024 9th International Conference on Intelligent Information Technology, this study by Albert C. Villaluz and colleagues bridges computer science and animal behavior to better understand canine emotions. The research highlights how artificial intelligence can enhance the interpretation of subtle non-verbal cues that dogs use to communicate with humans.
The team employed machine learning and computer vision techniques to detect and classify dogs’ emotional states. Using YOLOv5, a state-of-the-art object detection model, they identified dogs within video frames and still images. Emotional classification was achieved through transfer learning using the MobileNet architecture, allowing the model to differentiate among four categories: happiness, fear, anxiety, and aggressiveness.
Results revealed that the inclusion of facial features significantly improved accuracy compared to models based solely on body posture. The research also emphasized the critical role of balanced datasets and refined feature selection for reliable classification performance. These findings demonstrate that integrating multiple visual inputs—body movement, posture, and facial expression—enables AI to make more nuanced distinctions between emotional states.
This work represents a promising step toward the development of automated tools for emotion recognition in dogs. Potential applications include improving welfare monitoring in shelters, enhancing training environments, and supporting veterinarians and behaviorists in assessing emotional health. The study underscores the growing intersection of artificial intelligence and animal welfare science, where technology assists in decoding the emotional worlds of companion animals.
Source: Villaluz, A. C., Goma, J. C., Besa, J. V. T., Ignacio, J. I. D., & Zaguirre, S. A. A. (2024). Emotion Classification in Domestic Dogs Using Computer Vision Based on the Dog’s Body and Face. In Proceedings of the 2024 9th International Conference on Intelligent Information Technology. Published February 23, 2024.
