Abstract: This application provides an image scene recognition method and apparatus. The image scene recognition method includes: obtaining a to-be-recognized image, inputting the to-be-recognized image into a target visual element detection model and determining at least one target visual element comprised in the to-be-recognized image by the target visual element detection model, and inputting the at least one target visual element into a scene recognition model and generating a scene category corresponding to the to-be-recognized image by the scene recognition model. In this way, a bottom-up scene recognition method is used to infer, based on microscopic visual elements, a macroscopic scene category in which the microscopic visual elements may coexist, thereby implementing scene recognition, and greatly improving recognition precision and accuracy during image scene recognition.