AI-Generated Texture Image Dataset for Tactile Roughness Evaluation: Linguistic Prompt Ver

5.2 GB of texture images generated by a text-to-image AI model to evaluate tactile roughness perception. Yui Momiyama created this dataset to investigate whether visual features can quantify tactile roughness and how linguistic prompts influence the generated textures. The dataset, last updated in May 2026, includes images across wood, metal, stone, fiber, plastic, and ceramic categories.

Use Cases

• Training computer vision models to predict tactile roughness based on visual texture features mentioned in the description
• Studying the influence of linguistic descriptions on AI-generated material representations as described in the dataset's purpose
• Developing benchmarks for cross-modal perception linking vision and touch using the provided material category and roughness level labels
• Supporting reproducibility in research on AI-based material representation and visual texture analysis

Strengths

• Dataset size is explicitly 5.2 GB, indicating a substantial collection of images.
• Images are systematically generated across six defined material categories: wood, metal, stone, fiber, plastic, and ceramic.
• Includes controlled roughness levels and prompt information for reproducibility as stated in the description.

Limitations

• Row count (number of images) is unknown, which may limit suitability assessment for specific modeling tasks.
• Column-level documentation is absent; field semantics for metadata must be inferred after download.

Provenance

Source

figshare

Collection Method

Images generated using a text-to-image generative AI model with linguistic prompts controlling roughness.

Freshness

Last updated 2026-05-28 14:02:46; freshness should be verified.