Classification of Dog Emotion Using Transfer Learning on Convolutional Neural Network Algorithm

Recognizing your pet's emotions are very important to improve health, welfare and to detect certain diseases in the animal. The emotions in question are categorized into four categories, namely anger, happiness, calmness, and sadness. The model is designed by utilizing transfer learning techniques using the VGG16 architecture to perform image feature extraction for dog emotion classification based on the image of the animal's facial expression. The research produced an accuracy value of 96.72% on the training set and 88.05% on the validation set, as well as an average F1-Score value of 84.30% on the test set. This research shows the great potential of utilizing transfer learning in dog emotions classification and contributes to more advanced emotion recognition techniques to improve pet’s welfare.