'Exploiting Semantic Relations for Glass Surface Detection' (NeurIPS 2022)

[Project Page] [Paper] [Poster]

This repository is for the newly proposed dataset 'Glass Surface Detection - Semantics' ('GSD-S') along with our detection method 'Glass Semantic Network' (‘GlassSemNet’).

*GlassSemNetV2 is now available

Dataset Summary

Refer to utils/GSD-S_color_map.csv for semantic category color mapping.

Method Design

Comparison

Installation

Library dependencies require python 3.8+ environment and are summarized in requirements.txt.

> pip install -r requirements.txt

Demo

Inference and visualization scripts with respective required input directories. Trained model weights available for inference and testing.

> python predict.py -c CHECKPOINT -i IMAGE -o OUTPUT

> python visualize.py -i IMAGE -p PREDICTION -o OUTPUT

Evaluation script for performance assessment. Results by GlassSemNet available for reference.

> python evaluation.py -p PREDICTION -gt GROUNDTRUTH

Citation

@article{neurips2022:gsds2022,
  author    = {Lin, Jiaying and Yeung, Yuen-Hei and Lau, Rynson W.H.},
  title     = {Exploiting Semantic Relations for Glass Surface Detection},
  journal   = {NeurIPS},
  year      = {2022},
}