Concrete Aggregate Benchmark

The Concrete Aggregate Dataset consists of high resolution images acquired from 40 different concrete cylinders, cut lengthwise as to display the particle distribution in the concrete, with a ground sampling distance of 0.03mm. In order to train and evaluate approaches for the semantic segmentation of the concrete aggregate images, currently 17 of the 40 images have been annotated by manually associating one of the classes aggregate or suspension to each pixel. We encourage to use the remaining unlabelled images for semi-supervised segmentation approaches, in which unlabelled data is leveraged in addition to labelled training data in order to improve the segmentation performance.

In the subsequent figure, five examplary tiles of size 448x448 pixels and their annotated label masks are shown. The diversity of the appearance of both, aggregate and suspension can be noted.

In the figure below, the distribution of the aggregate particles in dependency on their sizes is depicted. The variation of the size of the particles contained in the data set ranges up to 15mm of maximum particle diameter. However, the majority of particles, namely more than 50% exhibit a maximum diameter of less then 3mm (100px). As a consequence, approximately 80% of the particles possess an area of 5mm ^2 or less.It has to be noted that particles with a size less then 20px are barely distinguishable from the suspension and are therefore not contained in the reference data.

If you make use of the proposed data, please cite the publication listed below.

Related Publications:

• Coenen, M.; Schack, T.; Beyer, D.; Heipke, C. and Haist, M. (2021): Semi-Supervised Segmentation of Concrete Aggregate using Consensus Regularisation and Prior Guidance. In: ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences V-2-2021, pp. 83–91, https://doi.org/10.5194/isprs-annals-V-2-2021-83-2021.