Active Multimodal Distillation for Few-shot Action Recognition

TL;DR

This paper introduces an active multimodal distillation framework for few-shot action recognition that dynamically identifies and leverages the most reliable modalities per sample, significantly improving recognition accuracy.

Contribution

It proposes a novel active inference-based approach to select reliable modalities and a mutual distillation module to enhance less reliable modality representations.

Findings

Outperforms existing methods on multiple benchmarks

Effectively identifies reliable modalities for each sample

Enhances representation learning through mutual distillation

Abstract

Owing to its rapid progress and broad application prospects, few-shot action recognition has attracted considerable interest. However, current methods are predominantly based on limited single-modal data, which does not fully exploit the potential of multimodal information. This paper presents a novel framework that actively identifies reliable modalities for each sample using task-specific contextual cues, thus significantly improving recognition performance. Our framework integrates an Active Sample Inference (ASI) module, which utilizes active inference to predict reliable modalities based on posterior distributions and subsequently organizes them accordingly. Unlike reinforcement learning, active inference replaces rewards with evidence-based preferences, making more stable predictions. Additionally, we introduce an active mutual distillation module that enhances the representation learning of less reliable modalities by transferring knowledge from more reliable ones. Adaptive multimodal inference is employed during the meta-test to assign higher weights to reliable modalities. Extensive experiments across multiple benchmarks demonstrate that our method significantly outperforms existing approaches.