Listening without Looking: Modality Bias in Audio-Visual Captioning

TL;DR

This paper investigates the reliance of audio-visual captioning models on sound versus vision, revealing a bias towards audio and proposing robustness tests and a new dataset to evaluate modality balance.

Contribution

It introduces systematic robustness tests for modality bias, and presents the AudioVisualCaps dataset to assess and improve modality balance in captioning models.

Findings

LAVCap shows a strong bias towards audio streams.

Training on AudioVisualCaps reduces modality bias.

Robustness tests reveal sensitivity to modality degradation.

Abstract

Audio-visual captioning aims to generate holistic scene descriptions by jointly modeling sound and vision. While recent methods have improved performance through sophisticated modality fusion, it remains unclear to what extent the two modalities are complementary in current audio-visual captioning models and how robust these models are when one modality is degraded. We address these questions by conducting systematic modality robustness tests on LAVCap, a state-of-the-art audio-visual captioning model, in which we selectively suppress or corrupt the audio or visual streams to quantify sensitivity and complementarity. The analysis reveals a pronounced bias toward the audio stream in LAVCap. To evaluate how balanced audio-visual captioning models are in their use of both modalities, we augment AudioCaps with textual annotations that jointly describe the audio and visual streams, yielding the AudioVisualCaps dataset. In our experiments, we report LAVCap baseline results on AudioVisualCaps. We also evaluate the model under modality robustness tests on AudioVisualCaps and the results indicate that LAVCap trained on AudioVisualCaps exhibits less modality bias than when trained on AudioCaps.