SuroTex: Surrounding Texture Dataset

Published: 16 September 2024| Version 1 | DOI: 10.17632/t6mdbpy2yk.1
Contributor:
Muhammad Ardi Putra

Description

This dataset contains 5000 texture images distributed across 50 classes with 100 images in each class. Despite being focused on the textures, color information (RGB) is still included in the 100 × 100 pixels images. As the name suggests, this dataset involves a variety of textures that can commonly be found in our surroundings, including textured surfaces from both indoor and outdoor areas.

Files

Steps to reproduce

The images for this dataset were captured using Samsung Galaxy Note 10 smartphone camera, in which it was configured to capture images at 9.1 megapixels (2268 × 4032). To generate the images, the first step to be done was to capture the textured surface. There were five images taken for every single texture class, where the raw JPG image was then resized into one-fourth of its original dimension prior to being cropped at random position 20 times with the cropping size of 100 × 100 pixels for each of the five images. This resizing process was done such that the repeating patterns are clearly seen even in the cropped images. The resulting images were then stored in separate folders according to their classes.

Institutions

Universitas Gadjah Mada

Categories

Computer Vision, Image Processing, Feature Extraction, Image Classification, Texture Analysis

Licence