OberflächenTex

OberflächenTex is a standardized dataset and benchmarking framework for the analysis of surface textures across materials and imaging modalities. The term combines the German words for surface and texture, signaling its focus on micro and mesoscale features visible on real-world materials. The project aims to support reproducible research by providing a common set of data, annotations, and evaluation protocols.

The collection includes high-resolution color and grayscale photographs of surfaces such as metals, ceramics, polymers, wood, stone, and composite materials. Samples are captured under varied illumination directions and viewing angles to reflect real-world imaging conditions. In addition to 2D images, the dataset contains 3D surface representations (depth maps or meshes) and, where available, multispectral or hyperspectral data.

Every sample is accompanied by metadata describing the material class, surface roughness, gloss level, color, porosity, microstructural features, and manufacturing or finishing processes. Imaging parameters include camera model, sensor resolution, distance, and lighting setup. Annotations cover texture classes, regions of interest, and, if measured, numerical roughness and gloss estimates.

OberflächenTex supports tasks including texture classification, material recognition, texture segmentation, roughness estimation, and 3D surface reconstruction. Evaluation protocols specify metrics such as classification accuracy, intersection-over-union, RMSE, and perceptual similarity scores.

The data are distributed under an open license suitable for research use, and provided in commonly used formats (PNG/JPEG, TIFF, JSON for metadata, and PLY for 3D). A central portal hosts downloads, documentation, and guidelines for contribution. Developed through international collaboration in the early 2020s, OberflächenTex has informed best practices in surface metrology and texture-aware machine learning, and it remains a reference point for benchmarking new methods.

See also: texture analysis, surface metrology, computer vision datasets.