Weakly-supervised Visual Instrument-playing Action Detection in Videos

TL;DR

This paper introduces a weakly-supervised visual method to detect when and where instruments are played in videos, leveraging auxiliary sound and object models to improve localization without extensive manual annotations.

Contribution

It presents a novel weakly-supervised framework that combines sound and object models to localize instrument-playing actions in videos, reducing the need for detailed annotations.

Findings

Significant improvement in localization accuracy

Effective use of auxiliary models for supervision

Validated on a manually annotated dataset

Abstract

Instrument playing is among the most common scenes in music-related videos, which represent nowadays one of the largest sources of online videos. In order to understand the instrument-playing scenes in the videos, it is important to know what instruments are played, when they are played, and where the playing actions occur in the scene. While audio-based recognition of instruments has been widely studied, the visual aspect of the music instrument playing remains largely unaddressed in the literature. One of the main obstacles is the difficulty in collecting annotated data of the action locations for training-based methods. To address this issue, we propose a weakly-supervised framework to find when and where the instruments are played in the videos. We propose to use two auxiliary models, a sound model and an object model, to provide supervisions for training the instrument-playing action model. The sound model provides temporal supervisions, while the object model provides spatial supervisions. They together can simultaneously provide temporal and spatial supervisions. The resulted model only needs to analyze the visual part of a music video to deduce which, when and where instruments are played. We found that the proposed method significantly improves the localization accuracy. We evaluate the result of the proposed method temporally and spatially on a small dataset (totally 5,400 frames) that we manually annotated.