Relation-Aware Distribution Representation Network for Person Clustering with Multiple Modalities

TL;DR

This paper introduces RAD-Net, a relation-aware distribution representation network that effectively captures multi-modal clues for person clustering, outperforming existing methods by leveraging relation-based features across modalities.

Contribution

The paper proposes a novel distribution representation for multi-modal clues that is modality agnostic, using a graph-based construction and cyclic refinement, improving clustering performance.

Findings

Achieved +6% F-score on VPCD dataset.

Achieved +8.2% F-score on VoxCeleb2 dataset.

Outperformed traditional multi-view clustering methods.

Abstract

Person clustering with multi-modal clues, including faces, bodies, and voices, is critical for various tasks, such as movie parsing and identity-based movie editing. Related methods such as multi-view clustering mainly project multi-modal features into a joint feature space. However, multi-modal clue features are usually rather weakly correlated due to the semantic gap from the modality-specific uniqueness. As a result, these methods are not suitable for person clustering. In this paper, we propose a Relation-Aware Distribution representation Network (RAD-Net) to generate a distribution representation for multi-modal clues. The distribution representation of a clue is a vector consisting of the relation between this clue and all other clues from all modalities, thus being modality agnostic and good for person clustering. Accordingly, we introduce a graph-based method to construct distribution representation and employ a cyclic update policy to refine distribution representation progressively. Our method achieves substantial improvements of +6% and +8.2% in F-score on the Video Person-Clustering Dataset (VPCD) and VoxCeleb2 multi-view clustering dataset, respectively. Codes will be released publicly upon acceptance.