DRSRD1: Sandstone and Carbonate Micro-CT Images for Super Resolution Training

Digital Rocks Super Resolution Dataset 1 contains organized 2D slices and 3D volumes of Bentheimer Sandstone and Estaillades Carbonate micro-CT images for machine learning. The dataset includes 800 images for training, 100 for validation, and 100 for testing, with high-resolution, 2x, and 4x downsampled versions. It was created by Ying Da Wang of UNSW Sydney and is structured similarly to the DIV2K benchmark dataset.

Use Cases

• Training super-resolution convolutional neural networks (SRCNN) based on the 2D and 3D micro-CT images.
• Benchmarking super-resolution algorithm performance on digital rock images, as described in the related publication.
• Validating image enhancement models using the provided high-resolution and downsampled image pairs.
• Analyzing rock microstructure properties using enhanced resolution images from the sandstone and carbonate samples.

Strengths

• Dataset is organized with 800 training, 100 validation, and 100 testing images, providing a clear split for model development.
• Includes both 2D (800x800 PNG) and 3D (80x80x80 MAT) data formats for different analysis needs.
• Provides multiple downsampled versions (2x and 4x) created with both default and randomized settings, useful for algorithm robustness testing.
• Contains data from two distinct rock types (Bentheimer Sandstone and Estaillades Carbonate) imaged at high resolutions of 3.8 and 3.1 microns.

Limitations

• Row count and total dataset size are unknown, which may limit suitability assessment for large-scale projects.
• Column-level documentation is absent; field semantics must be inferred after download.
• Last update date is unknown; freshness unverified.

Provenance

Source

UNSW Sydney (Bentheimer Sandstone) and Digital Rocks Portal (Estaillades Carbonate).

Collection Method

Micro-CT imaging and subsequent cropping and downsampling using Matlab functions.