The current image-text cross-modal retrieval faces challenges due to the heterogeneous nature of different modalities. This paper proposes an improved model based on multi-scale modal distance ...