# Glass Segmentation with RGB-Thermal Image Pairs

Source: [Glass Segmentation with RGB-Thermal Image Pairs](<https://www.sota2.com/research/paper/68fa6ec61e209131cea41f9f>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-01-04T03:41:25.908Z

Authors: Dong Huo, Jian Wang, Yiming Qian, Yee-Hong Yang

Published: 2022-04-12T00:20:22.000Z

## Abstract

This paper proposes a new glass segmentation method utilizing paired RGB and thermal images. Due to the large difference between the transmission property of visible light and that of the thermal energy through the glass where most glass is transparent to the visible light but opaque to thermal energy, glass regions of a scene are made more distinguishable with a pair of RGB and thermal images than solely with an RGB image. To exploit such a unique property, we propose a neural network architecture that effectively combines an RGB-thermal image pair with a new multi-modal fusion module based on attention, and integrate CNN and transformer to extract local features and non-local dependencies, respectively. As well, we have collected a new dataset containing 5551 RGB-thermal image pairs with ground-truth segmentation annotations. The qualitative and quantitative evaluations demonstrate the effectiveness of the proposed approach on fusing RGB and thermal data for glass segmentation. Our code and data are available at https://github.com/Dong-Huo/RGB-T-Glass-Segmentation.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

- [Glass Segmentation on RGB-T dataset Images with glass 1.0 (test)](<https://www.sota2.com/research/sota/glass-segmentation-on-rgb-t-dataset-images-with-glass-1-0-test.md>): Glass Segmentation; RGB-T dataset Images with glass 1.0 (test); MAE: 2.7.

- [Glass Segmentation on RGB-T dataset Images without glass 1.0 (test)](<https://www.sota2.com/research/sota/glass-segmentation-on-rgb-t-dataset-images-without-glass-1-0-test.md>): Glass Segmentation; RGB-T dataset Images without glass 1.0 (test); MAE: 0.003.

- [Glass Segmentation on RGB-T dataset All images 1.0 (test)](<https://www.sota2.com/research/sota/glass-segmentation-on-rgb-t-dataset-all-images-1-0-test.md>): Glass Segmentation; RGB-T dataset All images 1.0 (test); MAE: 0.024.

- [Salient Object Detection on VT1000 47](<https://www.sota2.com/research/sota/salient-object-detection-on-vt1000-47.md>): Salient Object Detection; VT1000 47; MAE: 2.

- [Glass Segmentation on GDD 4 (test)](<https://www.sota2.com/research/sota/glass-segmentation-on-gdd-4-test.md>): Glass Segmentation; GDD 4 (test); MAE: 0.072.

- [Salient Object Detection on VT5000 122](<https://www.sota2.com/research/sota/salient-object-detection-on-vt5000-122.md>): Salient Object Detection; VT5000 122; MAE: 0.036.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2204.05453>)

[PDF](<https://arxiv.org/pdf/2204.05453.pdf>)

[Implementation](<https://github.com/Dong-Huo/RGB-T-Glass-Segmentation>)