With the rapid development of artificial intelligence technologies such as deep learning and machine learning, a large number of datasets have been generated in the field of remote sensing to meet the diverse needs of remote sensing tasks. However, these datasets exhibit significant differences in type, size, spatial and temporal resolution. To address this issue, this paper proposes a standardized multimodal sample dataset organization method oriented towards artificial intelligence interpretation. This establishes an effective link between samples and image data, forming a temporal sample dataset based on geographical spatial locations. This meets the demands of intelligent interpretation for a data cube that is scalable in terms of big data, multimodality, and samples. It achieves orderly management of massive remote sensing sample data while also supporting continuous expansion and reuse of samples. This paper applies the multimodal sample dataset organization method from both dataset and systematic management, providing effective data and technical support for artificial intelligence interpretation in various remote sensing tasks.