Causalverse Image

CausalVerse Image Dataset contains two families of splits: Physics splits (Fall, Refraction, Slope, Spring) and Static image generation splits (scene1-4). The dataset is associated with the paper 'CausalVerse: Benchmarking Causal Representation Learning with Configurable High-Fidelity Simulations' and was last updated on 2025-10-23.

Use Cases

• Benchmarking causal representation learning models based on configurable physics simulations.
• Training models to infer causal relationships from visual scenes based on the described physics splits.
• Evaluating model performance on static image generation tasks based on the scene splits.
• Developing methods for extracting structured metadata from images based on the per-sample 'metavalue' field.

Strengths

• Dataset is structured into distinct splits for physics and static image generation, allowing for targeted experimentation.
• All splits share a consistent schema with image, path, and metadata columns, which likely aids in processing.
• Associated with a named research paper, providing academic context and validation.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Description metadata is limited; actual data quality requires manual inspection after download.

Provenance

Source

CausalVerse

Collection Method

Configurable high-fidelity simulations, as described in the associated paper.

Freshness

Last updated 2025-10-23 10:42:00.