Paper Image

Generating text to improve image classification

Published on:

21 July 2023

Primary Category:

Computer Vision and Pattern Recognition

Paper Authors:

Kathleen M. Lewis,

Emily Mu,

Adrian V. Dalca,

John Guttag


Key Details

Method generates descriptive text for fine-grained image classes

Matches images to label-preserving text about visual features

Fine-tunes vision-language model on image-text pairs

Improves classification over vision-only and vision-language baselines

Works well for full-shot and few-shot classification tasks

AI generated summary

Generating text to improve image classification

This paper proposes a method to generate descriptive text for images in fine-grained classification datasets. The method uses a large language model to create detailed text about visual features for each class. It then matches images to label-preserving text. Fine-tuning a vision-language model on the image-text pairs improves classification.

Answers from this paper


No comments yet, be the first to start the conversation...

Sign up to comment on this paper

Sign Up