Toward Verifiable and Reproducible Human Evaluation for Text-to-Image Generation

Human evaluation is critical for validating the performance of text-to-image generative models, as this highly cognitive process requires deep comprehension of text and images. However, our survey of 37 recent papers reveals that many works rely solely on automatic measures (e.g., FID) or perform po...

Full description

Saved in:
Bibliographic Details
Published in:arXiv.org 2023-04
Main Authors: Otani, Mayu, Togashi, Riku, Sawai, Yu, Ishigami, Ryosuke, Nakashima, Yuta, Rahtu, Esa, Heikkilä, Janne, Satoh, Shin'ichi
Format: Article
Language:English
Subjects:
Human performance
Image processing
Performance evaluation
Reproducibility
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Human evaluation is critical for validating the performance of text-to-image generative models, as this highly cognitive process requires deep comprehension of text and images. However, our survey of 37 recent papers reveals that many works rely solely on automatic measures (e.g., FID) or perform poorly described human evaluations that are not reliable or repeatable. This paper proposes a standardized and well-defined human evaluation protocol to facilitate verifiable and reproducible human evaluation in future works. In our pilot data collection, we experimentally show that the current automatic measures are incompatible with human perception in evaluating the performance of the text-to-image generation results. Furthermore, we provide insights for designing human evaluation experiments reliably and conclusively. Finally, we make several resources publicly available to the community to facilitate easy and fast implementations.
ISSN:2331-8422