Abstract
This data article describes a dataset of images of common Chinese deities. The dataset is divided into five categories according to the types of deities, and a total of 1314 original images were captured by smart phones from Chinese temples and through Google search engine. Each category were split into training, validation and test subsets in a ratio of 70:20:10. We rotated the pictures by 30°, 60°, 90°, 120°, 150°, and 180°; and zoomed in and out to augment the images for training and validation sets. After data enhancement, the total number of images reaches 10,786. Two models, EfficientNet-B0 and MobileNetV2, are used to identify five kinds of god images. After data augmentation, the accuracy, precision, recall, specificity and F1-score of EfficientNet-B0 were 96.15%, 96.44%, 96.18%, 96.16% and 97.60%, respectively; the accuracy, precision recall, specificity and F1-score of MobileNetV2 were 92.31%, 92.89%, 92.37%, 92.33% and 95.19%, respectively. This dataset can be used as a reference for traditional Chinese god statue images, and can also be used for object detection and image classification through machine learning and deep learning methods.