Furniture Catalog is a dataset of 5,171 individual furniture images across two room types and 1,781 room scene images, created by Darebal for a bachelor's thesis at Ukrainian Catholic University. It includes triplet training data (golden/train/val splits) used to train a visual furnishings compatibility model. The dataset was last updated on May 2, 2026.
Use Cases
- Train visual compatibility models based on the triplet training splits mentioned in the description.
- Benchmark image retrieval systems for furniture based on the catalog of individual furniture images.
- Generate synthetic room scenes based on the collection of 1,781 room scene images.
- Study machine learning applications in interior design based on the dataset's stated purpose.
Strengths
- Contains 5,171 individual furniture images.
- Includes 1,781 room scene images.
- Provides triplet training data with golden, train, and validation splits.
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
- Darebal (author), likely associated with Ukrainian Catholic University.
- Collection Method
- Created for the bachelor's thesis "Visual Furnishings Compatibility Learning and Retrieval Using Machine Learning".
- Freshness
- Last updated 2026-05-02 19:15:27; freshness should be verified.