BiMa: Towards Biases Mitigation for Text-Video Retrieval via Scene Element Guidance

TL;DR

BiMa is a framework that reduces biases in text-video retrieval by incorporating scene elements into video representations and disentangling textual content from biases, improving performance and robustness.

Contribution

This paper introduces BiMa, a novel bias mitigation framework for TVR that leverages scene element guidance and feature disentanglement, a new approach in the field.

Findings

BiMa improves retrieval accuracy across five benchmarks.

The model effectively reduces biases in out-of-distribution tasks.

Extensive ablations validate the components' effectiveness.

Abstract

Text-video retrieval (TVR) systems often suffer from visual-linguistic biases present in datasets, which cause pre-trained vision-language models to overlook key details. To address this, we propose BiMa, a novel framework designed to mitigate biases in both visual and textual representations. Our approach begins by generating scene elements that characterize each video by identifying relevant entities/objects and activities. For visual debiasing, we integrate these scene elements into the video embeddings, enhancing them to emphasize fine-grained and salient details. For textual debiasing, we introduce a mechanism to disentangle text features into content and bias components, enabling the model to focus on meaningful content while separately handling biased information. Extensive experiments and ablation studies across five major TVR benchmarks (i.e., MSR-VTT, MSVD, LSMDC, ActivityNet, and DiDeMo) demonstrate the competitive performance of BiMa. Additionally, the model's bias mitigation capability is consistently validated by its strong results on out-of-distribution retrieval tasks.