Predicting Pet Photo Pawpularity: A Deep Learning Approach on PetFinder.my

PetFinder.my, a leading pet adoption platform, plays a crucial role in matching rescue animals with potential adopters. In this context, the visual appeal of pet photographs, termed “Pawpularity,” significantly influences adoption rates. This study harnesses advanced deep learning techniques to predict and enhance the Pawpularity of pet photos, aiming to boost adoption success. By utilizing a class of convolutional neural network, EfficientNet, the research integrates image features with meta-data to predict the Pawpularity scores of pet photos. Opti-mized through iterative training and validation, the model's architecture leverages pretrained networks and fine-tuning to improve prediction accuracy and reduce the loss. Preliminary results indicate that the model can successfully assess the Paw-pularity of pet images, offering insights into how specific photo characteristics affect viewer appeal. By systematically evaluating and enhancing the attractiveness of pet photos, this research contributes to increasing adoption rates, thereby advancing the welfare of rescue animals. Furthermore, the integration of image analysis and metadata evaluation in a deep learning framework presents a novel approach to understanding and enhancing the factors contributing to successful pet adoptions on PetFinder.my. However the proposed deep learning model attain the lowest RMSE of 18.9648 during model training.