Paper Image

Text and Masks Generate Segmentation Data

Published on:

6 November 2023

Primary Category:

Computer Vision and Pattern Recognition

Paper Authors:

Hanrong Ye,

Jason Kuen,

Qing Liu,

Zhe Lin,

Brian Price,

Dan Xu


Key Details

SegGen generates masks from text prompts and images from masks

Two approaches proposed: MaskSyn for new masks and ImgSyn for new images

Strongly boosts leading segmentation models on ADE20K, COCO, without extra data

Enhances model robustness and generalization to unfamiliar images

Establishes new state-of-the-art results on highly competitive benchmarks

AI generated summary

Text and Masks Generate Segmentation Data

This paper proposes a new method called SegGen to generate high-quality training data for image segmentation models. It uses text prompts to generate diverse segmentation masks, then generates corresponding aligned images. This synthetic data significantly boosts performance.

Answers from this paper


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

Sign up to comment on this paper

Sign Up