Grounding Object Detections With Transcriptions

TL;DR

This paper introduces a method to automatically generate labeled entity-video frame pairs from instruction videos using speech transcriptions, facilitating supervised learning without manual annotation.

Contribution

It presents a novel approach to automatically extract training data from videos and transcriptions, enabling scalable dataset creation for visual grounding tasks.

Findings

Effective extraction of entity-video pairs demonstrated

Models trained on automatically generated data perform well

First step towards automated dataset construction for vision-language tasks

Abstract

A vast amount of audio-visual data is available on the Internet thanks to video streaming services, to which users upload their content. However, there are difficulties in exploiting available data for supervised statistical models due to the lack of labels. Unfortunately, generating labels for such amount of data through human annotation can be expensive, time-consuming and prone to annotation errors. In this paper, we propose a method to automatically extract entity-video frame pairs from a collection of instruction videos by using speech transcriptions and videos. We conduct experiments on image recognition and visual grounding tasks on the automatically constructed entity-video frame dataset of How2. The models will be evaluated on new manually annotated portion of How2 dev5 and val set and on the Flickr30k dataset. This work constitutes a first step towards meta-algorithms capable of automatically construct task-specific training sets.