Interpretable Motion-Attentive Maps: Spatio-Temporally Localizing Concepts in Video Diffusion Transformers

TL;DR

This paper proposes a motion-feature selection algorithm to obtain an Interpretable Motion-Attentive Map (IMAP) that localizes motion spatially and temporally and introduces GramCol, which adaptively produces per-frame saliency maps for any text concept, including both motion and non-motion.

Abstract

Video Diffusion Transformers (DiTs) have been synthesizing high-quality video with high fidelity from given text descriptions involving motion. However, understanding how Video DiTs convert motion words into video remains insufficient. Furthermore, while prior studies on interpretable saliency maps primarily target objects, motion-related behavior in Video DiTs remains largely unexplored. In this paper, we investigate concrete motion features that specify when and which object moves for a given motion concept. First, to spatially localize, we introduce GramCol, which adaptively produces per-frame saliency maps for any text concept, including both motion and non-motion. Second, we propose a motion-feature selection algorithm to obtain an Interpretable Motion-Attentive Map (IMAP) that localizes motion spatially and temporally. Our method discovers concept saliency maps without the need for any gradient calculation or parameter update. Experimentally, our method shows outstanding localization capability on the motion localization task and zero-shot video semantic segmentation, providing interpretable and clearer saliency maps for both motion and non-motion concepts.

Figures (30)

• Figure 1: Interpretable Motion-Attentive Maps (IMAP). IMAP is an interpretable map that spatially and temporally localizes any motion concept (e.g., lightning strike). Without additional training, IMAP is obtained from the features of video diffusion transformers at motion-specific attention heads.
• Figure 2: Spatiotemporal localization pipeline. This pipeline obtains a video saliency map for any concept using Video DiTs. Given a concept, we first obtain a text-surrogate token via Query–Key Matching, and then compute the GramCol to derive its spatial saliency map. For motion concepts, we additionally identify motion heads before computing GramCol , thereby improving temporal localization.
• Figure 3: Layer-wise average of $\lambda_2$ and the corresponding feature maps. As the average $\lambda_2$ increases, the extracted features become sharper and more interpretable.
• Figure 4: QK-Matching visualizations per text token. While QK-Matching yields somewhat unclear spatial localization, its peak (red dot) still accurately pinpoints the target concept.
• Figure 5: Visualization of Motion Localization Score (MLS) versus the separation score (Calinski-Harabasz index, CHI). MLS measured from GramCol extracted across attention heads in layers $\mathcal{L}$ of CogVideoX-2B. Heads with higher CHI scores tend to exhibit higher MLS, with a Pearson correlation coefficient of 0.60.