Sign in to view source links and access this dataset
Description
Concept-Targeted Causal Images is a concept-centric dataset designed for studying causal visual representations in the brain. The dataset, created by BrainCause, contains three complementary image types for each concept: positive depictions, semantic negatives, and counterfactual edits. It was last updated on May 22, —.
Use Cases
Training models for causal visual concept disentanglement based on the three image types.
Benchmarking neural network interpretability methods using the counterfactual edits.
Studying brain-inspired visual representation learning based on the semantic negative images.
Strengths
Dataset is structured around a specific research goal: studying causal visual representations.
Provides three distinct, complementary image types (positive, semantic negative, counterfactual) per concept.
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
BrainCause
Collection Method
Likely curated and edited for research purposes, as indicated by the description of counterfactual edits.
Time Range
null
Freshness
Last updated 2026-05-22 12:45:44; freshness should be verified.
Geography
null
License is unknown; terms of use must be verified before application.