Synthetic Texture Images for CNN Analysis in Spatial and Fourier Domains

2.5 MB of pre-trained deep learning weights, configuration code, and evaluation scripts for synthetic texture analysis. The dataset, associated with a manuscript submitted to The Visual Computer, is authored by Eduardo Bianucci Barcelona and was last updated on 2026-05-28. Full raw image partitions for training and testing are hosted separately on GitHub due to file size constraints.

Use Cases

• Training convolutional neural networks for texture classification based on the described synthetic textures.
• Comparing model performance in spatial versus Fourier domains as indicated by the project title.
• Reproducing or extending the experimental results from the associated academic manuscript.
• Benchmarking CNN architectures on controlled synthetic image data.

Strengths

• Includes pre-trained model weights, enabling immediate evaluation and transfer learning.
• Associated code and scripts are provided, facilitating reproducibility of the research.
• Has a permanent DOI (10.6084/m9.figshare.XXXXXXX) for versioned citation.

Limitations

• The primary image datasets (training_set.zip, test_set.zip) are not directly included in this 2.5 MB package; they must be downloaded separately from GitHub.
• Row and column counts for the image data are unknown, limiting initial assessment of dataset scale and structure.
• Description metadata is limited; actual data quality and image characteristics require manual inspection after download.

Provenance

Source

Eduardo Bianucci Barcelona

Collection Method

Likely generated synthetically for controlled computer vision experiments.

Freshness

Last updated 2026-05-28 20:59:47; freshness should be verified.