A study on dog emotion recognition using deep learning techniques

• Main Authors: John, Sarah Merin, Nagaraj, Ranjith Kumar Kenchinapalya, Lobo, Clevan Cleon, Subramani, Sivakannan
• Format: Conference Proceeding
• Language: English
• Subjects: Deep learning; Dogs; Emotion recognition; Emotions; Machine learning
• ISSN: 0094-243X; 1551-7616
• DOI: 10.1063/5.0152413

The dog (Canis familiaris) is an exceptional creature and one of the most seasoned known trained species. Canines have existed together with people since they were trained something like 30,000years prior. They have great socio-cognitive abilities as well as exceptional interpersonal capabilities. They use facial expressions, actions, gestures and behavior to convey their emotions. These expressions are important for social communication since nonverbal communication relies heavily on facial expressions. They are fairly accurate indicators of emotions. Although much research has been conducted on emotions in humans, we relatively know very little about understanding animal emotions. Identifying these emotional expressions is important yet little or no research has been carried on this. In dogs, for instance, the study is crucial not simply to address fundamental and applied logical worries, however for functional contemplations, because many undesirable behaviors, such as aggressiveness based on frustration, are considered to have an emotional foundation. The objective of this paper is to see if a deep learning system can recognize different facial emotion classes in Dogs such as happiness, sadness, anger, and neutral. We trained and analyzed three different convolutional networks for this: a shallow CNN, sequential CNN, and MBCC-CNN. Following the experiments conducted using the above-mentioned architectures we obtained an accuracy of 58%, 61.28% and 71.26% respectively. The results indicate that MBCC-CNN can perform better than the other two models. Subsequently, this paper proposes a strategy dependent on various Multi branch cross-connectional convolutional neural net (MBCC-CNN) for perceiving feelings in canines.