De La Salle University – Outdoor Mirrors and Reflective Surfaces (DLSU-OMRS) Dataset

<p><strong>The De La Salle University &ndash; Outdoor Mirrors and Reflective Surfaces (DLSU-OMRS) dataset contains&nbsp;454 images of outdoor mirrors and reflective surfaces, along with their corresponding ground-truth masks for segmentation</strong>. The images were&nbsp;scraped from Shutterstock using the key&nbsp;phrases <em>outdoor mirror</em>&nbsp;and <em>street mirror</em>&nbsp;and manually&nbsp;filtered to remove duplicates and heavily manipulated&nbsp;photos.&nbsp;Ground-truth masks were produced through manual segmentation.</p> <p>The images have their respective licenses, and the ground-truth masks are licensed under the BSD 3-Clause &quot;New&quot; or &quot;Revised&quot; License. The use of this dataset is restricted to noncommercial purposes only.</p> <p>More details can be found in the paper&nbsp;&quot;<strong>Designing a Lightweight Edge-Guided Convolutional Neural Network for Segmenting Mirrors and Reflective Surfaces</strong>,&quot; which was&nbsp;accepted for full paper presentation at the <strong>2023 International Conference in Central Europe on Computer Graphics, Visualization and Computer Vision (WSCG 2023)</strong>.&nbsp;The project page is&nbsp;<a href="https://github.com/memgonzales/mirror-segmentation">https://github.com/memgonzales/mirror-segmentation</a>. The paper is published in&nbsp;<em>Computer Science Research Notes</em>:&nbsp;<a href="http://wscg.zcu.cz/WSCG2023/full/E59-full.pdf">http://wscg.zcu.cz/WSCG2023/full/E59-full.pdf</a>.</p>