Abstract
Texture analysis can be considered as one of the most important topics in the field of image processing and computer vision. However, the existing texture datasets such as KTH-TIPS, KTH-TIPS2, USPTex, DTD, and ALOT still have limitations which causes the resulting analysis on different texture classification algorithms to be somewhat unreliable. The two main reasons behind this problem are the limited number of texture classes and the non-uniformity of the image sizes. To address this issue, we introduce a new texture classification dataset named SuroTex (Surrounding Texture) Dataset. The dataset, which was collected back in November 2023 in Ulsan, South Korea, contains 5000 image textures. These number of samples are divided into 50 classes, each with 100 RGB samples of size 400 × 400 pixels. Future researchers can use SuroTex for a benchmark dataset both for general image classification or specifically for texture classification tasks. Additionally, models trained on this dataset can also act as a pre-trained model which can further be fine-tuned for similar purposes such as texture-based material classification.