DORi: Discovering Object Relationship for Moment Localization of a Natural-Language Query in Video

TL;DR

This paper introduces DORi, a novel approach for localizing specific moments in videos based on natural language queries by learning a language-conditioned video embedding that captures object, human, and activity relationships.

Contribution

The paper proposes a new message-passing algorithm that models spatial and temporal relationships in videos conditioned on language queries for improved moment localization.

Findings

Outperforms state-of-the-art on three benchmark datasets

Introduces YouCookII as a new benchmark dataset

Effective in capturing complex object-human-activity relationships

Abstract

This paper studies the task of temporal moment localization in a long untrimmed video using natural language query. Given a query sentence, the goal is to determine the start and end of the relevant segment within the video. Our key innovation is to learn a video feature embedding through a language-conditioned message-passing algorithm suitable for temporal moment localization which captures the relationships between humans, objects and activities in the video. These relationships are obtained by a spatial sub-graph that contextualizes the scene representation using detected objects and human features conditioned in the language query. Moreover, a temporal sub-graph captures the activities within the video through time. Our method is evaluated on three standard benchmark datasets, and we also introduce YouCookII as a new benchmark for this task. Experiments show our method outperforms state-of-the-art methods on these datasets, confirming the effectiveness of our approach.