Multilevel Language and Vision Integration for Text-to-Clip Retrieval

TLDR

The paper tackles text‑based activity retrieval in video. The goal is to retrieve video clips that match a given activity description. The authors propose a multilevel model that injects text features early during clip proposal generation, uses visual features to modulate word‑level query processing in an RNN, and incorporates a multi‑task loss with query re‑generation. The method outperforms previous approaches on Charades‑STA and ActivityNet Captions.

Abstract

We address the problem of text-based activity retrieval in video. Given a sentence describing an activity, our task is to retrieve matching clips from an untrimmed video. To capture the inherent structures present in both text and video, we introduce a multilevel model that integrates vision and language features earlier and more tightly than prior work. First, we inject text features early on when generating clip proposals, to help eliminate unlikely clips and thus speed up processing and boost performance. Second, to learn a fine-grained similarity metric for retrieval, we use visual features to modulate the processing of query sentences at the word level in a recurrent neural network. A multi-task loss is also employed by adding query re-generation as an auxiliary task. Our approach significantly outperforms prior work on two challenging benchmarks: Charades-STA and ActivityNet Captions.