SCDM: Spatial-Contextual Disentanglement Mamba via Differential Inference for Efficient Image Classification
The episode discusses the paper "SCDM: Spatial-Contextual Disentanglement Mamba via Differential Inference for Efficient Image Classification." Hosts analyze how SCDM uses a dual-branch structure to separate pathological features from normal anatomy in sequence models, focusing on its mechanism, efficiency gains, and improved spatial selectivity for medical image classification.