Annotation Methodologies for Vision and Language Dataset Creation

TL;DR

This paper discusses the challenges and issues encountered during the creation and validation of annotated datasets for vision and language tasks, highlighting common problems in data selection and annotation processes.

Contribution

It provides an analysis of difficulties faced in dataset creation for vision-language tasks, emphasizing the need for improved annotation methodologies.

Findings

Identifies common problems in dataset annotation processes

Highlights issues in data validation for vision-language datasets

Suggests areas for improving annotation methodologies

Abstract

Annotated datasets are commonly used in the training and evaluation of tasks involving natural language and vision (image description generation, action recognition and visual question answering). However, many of the existing datasets reflect problems that emerge in the process of data selection and annotation. Here we point out some of the difficulties and problems one confronts when creating and validating annotated vision and language datasets.