Missing Modality

Missing modality in multimodal learning addresses the challenge of incomplete data where some input sources (modalities) are unavailable during model training or inference. Current research focuses on developing robust models that can effectively utilize available modalities to compensate for missing ones, employing techniques like masked modality projection, modality-aware prompt learning, and various data augmentation strategies within diverse architectures including transformers and federated learning frameworks. This research is crucial for improving the reliability and generalizability of multimodal systems across various applications, particularly in scenarios with inherent data scarcity or limitations in data acquisition, such as healthcare and autonomous driving.