ShotcreteDepth: Bi-modal Images for Robotic Depth Perception in Construction

ShotcreteDepth captures construction environments for developing stereo matching, depth completion, and depth estimation methods. The 40.2 GB dataset is released by author Jakub Gregorek under a CC-BY-SA-4.0 license. More details are available in the associated GitHub repository.

Use Cases

• Develop stereo matching methods based on bi-modal imagery of shotcrete environments.
• Train depth completion models based on data from construction sites.
• Evaluate depth estimation algorithms based on challenging real-world construction scenes.

Strengths

• Dataset size is 40.2 GB, indicating substantial data volume.
• Explicitly designed for robust depth perception in challenging, unstructured environments.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.

Provenance

Source

Jakub Gregorek

Freshness

Last updated 2026-05-12 09:01:33; freshness should be verified.