Multi-Modal Vision Transformers for Crop Mapping from Satellite Image Time Series
Using images acquired by different satellite sensors has shown to improve classification performance in the frame workofcropmappingfromsatellite image time series (SITS). Existing state-of-the-art architectures use self-attention mech anisms to process the temporal dimension and convolutions for the spatial dimension of SITS. Motivated by the success of purely attention-based architectures in crop mapping from single-modal SITS, we introduce several multi-modal multi temporal transformer-based architectures. Specifically, we investigate the effectiveness of Early Fusion, Cross Atten tion Fusion and Synchronized Class Token Fusion within the Temporo-Spatial Vision Transformer (TSViT). Experi mental results demonstrate significant improvements over state-of-the-art architectures with both convolutional and self-attention components. Index Terms— Multi-modal fusion, time series classifi cation, crop mapping, transformers, remote sensing.
