Beyond Uncertainty: Evidential Deep Learning for Robust Video Temporal Grounding

TL;DR

This paper introduces SRAM, a novel network module for Video Temporal Grounding that incorporates Deep Evidential Regression to explicitly quantify uncertainty, improving robustness and interpretability in open-world scenarios.

Contribution

It presents the first successful application of Deep Evidential Regression in Video Temporal Grounding, with a new Geom-regularizer to enhance uncertainty estimation.

Findings

Improved robustness to noisy and out-of-distribution data.

Enhanced interpretability through explicit uncertainty quantification.

Demonstrated effectiveness on extensive experiments.

Abstract

Existing Video Temporal Grounding (VTG) models excel in accuracy but often overlook open-world challenges posed by open-vocabulary queries and untrimmed videos. This leads to unreliable predictions for noisy, corrupted, and out-of-distribution data. Adapting VTG models to dynamically estimate uncertainties based on user input can address this issue. To this end, we introduce SRAM, a robust network module that benefits from a two-stage cross-modal alignment task. More importantly, it integrates Deep Evidential Regression (DER) to explicitly and thoroughly quantify uncertainty during training, thus allowing the model to say "I do not know" in scenarios beyond its handling capacity. However, the direct application of traditional DER theory and its regularizer reveals structural flaws, leading to unintended constraints in VTG tasks. In response, we develop a simple yet effective Geom-regularizer that enhances the uncertainty learning framework from the ground up. To the best of our knowledge, this marks the first successful attempt of DER in VTG. Our extensive quantitative and qualitative results affirm the effectiveness, robustness, and interpretability of our modules and the uncertainty learning paradigm in VTG tasks. The code will be made available.