---
license: cc-by-nc-4.0
tags:
  - furniture
  - interior-design
  - image-retrieval
  - compatibility
  - computer-vision
task_categories:
  - image-to-image
---

# Furniture Catalog

Image catalog and training splits from the bachelor's thesis **"Visual Furnishings Compatibility Learning and Retrieval Using Machine Learning"** (Ukrainian Catholic University, 2026).

5 171 individual furniture images across two room types, 1 781 room scene images, plus triplet training data (golden / train / val splits) used to train the compatibility model.

## Repo layout

```
{room}/
  {category}/           *.jpg   — individual furniture images
  general_manifest.json         — scene metadata + product-page links
  triplets/
    scene_split.json            — golden(18%) / train(70%) / val(12%) assignment
    train_triplets.csv
    val_triplets.csv
    golden_triplets.csv
  embeddings/
    furniture_embeddings.json   — ResNet50/CLIP features used for negative mining
scenes/bedrooms/
  deepfurn/{scene}/scene_image.jpg
  sklad_mebliv/{scene}/scene_image.jpg
scenes/living_rooms/
  deepfurn/{scene}/scene_image.jpg
```

Triplet CSV columns: `anchor_id`, `anchor_category`, `anchor_scene`, `anchor_source`,
`positive_id`, `positive_category`, `positive_scene`, `positive_source`,
`negative_id`, `negative_category`, `negative_scene`, `negative_source`,
`pos_neg_distance`, `anchor_counterpart_distance`, `combined_distance`.

## Data sources

- **DeepFurniture** — Liu et al., "Furnishing Your Room by What You See", arXiv:1911.09299.
  Original dataset: [byliu/DeepFurniture](https://huggingface.co/datasets/byliu/DeepFurniture)
- **Sklad Mebliv** — Ukrainian furniture retailer: [skladmebliv.ua](https://skladmebliv.ua/)

Images are used for non-commercial academic research only. If you are a rights holder with concerns, please open an issue.

## Quick start

```bash
git clone https://github.com/Baredal/thesis_furnitures
cd thesis_furnitures
pip install -r requirements.txt
python download_from_hf.py        # downloads this dataset + the model repo
streamlit run src/app/streamlit_app.py
```

## Related

- **Model weights + retrieval artifacts**: [Darebal/furniture-compatibility-siamese](https://huggingface.co/Darebal/furniture-compatibility-siamese)
- **Code**: [github.com/Baredal/thesis_furnitures](https://github.com/Baredal/thesis_furnitures)

## Citation

```bibtex
@thesis{Baredal2026furniture,
  title  = {Visual Furnishings Compatibility Learning and Retrieval Using Machine Learning},
  author = {Yaroslav-Dmytro Strus},
  school = {Ukrainian Catholic University},
  year   = {2026},
  type   = {Bachelor's Thesis}
}
```

Please also cite DeepFurniture if you use this dataset:

```bibtex
@article{liu2019furnishing,
  title   = {Furnishing Your Room by What You See: An End-to-End Furniture Set Retrieval Framework with Rich Annotated Benchmark Dataset},
  author  = {Bingyuan Liu and Jiantao Zhang and Xiaoting Zhang and Wei Zhang and Chuanhui Yu and Yuan Zhou},
  journal = {arXiv preprint arXiv:1911.09299},
  year    = {2019},
  url     = {https://arxiv.org/abs/1911.09299}
}
```

## License

Derived artefacts (triplet CSVs, manifests): CC BY-NC 4.0.  
Original images remain under the terms of their respective sources — [DeepFurniture](https://huggingface.co/datasets/byliu/DeepFurniture) and [Sklad Mebliv](https://skladmebliv.ua/) — and are used for non-commercial academic research only.