README.md

---
license: cc-by-2.0
language:
- en
- nl
pretty_name: ArtiFact
task_categories:
- image-classification
- text-classification
- question-answering
- object-detection
tags:
- art
- museum
- multimodal
- data-cleaning
- error-detection
- cultural-heritage
- table
- image
- text
size_categories:
- 100K<n<1M
---


# ArtiFact

[ArtiFact](https://olgaovcharenko.github.io/ArtiFact/) is a large-scale multimodal benchmark of museum artwork records with aligned images and structured metadata. It is designed for evaluating metadata extraction, error detection, semantic querying, and multimodal reasoning over cultural-heritage collections.

The dataset combines records from the **Rijksmuseum**, the **Metropolitan Museum of Art (Met)**, and the **Art Institute of Chicago (AIC)**, with normalized fields for artists, dates, materials, techniques, dimensions, culture, location, and descriptions.

<img src="fig1_overview.png" alt="" style="display: inline-block; height: auto; width: auto; vertical-align: text-bottom; margin: 0 0.0rem;" />

## Dataset structure

The Hub repository is organized as follows:

```
data/
├── ArtiFact_clean.csv
├── ArtiFact_clean_dirty.csv
├── ArtiFact_clean_dirty_sample_10.csv
├── ArtiFact_clean_dirty_sample_200.csv
├── ArtiFact_clean_dirty_sample_500.csv
└── images/
    ├── ArtiFact_clean.tar              # all clean-split images
    ├── ArtiFact_clean_dirty.tar        # all dirty-split images
    └── ArtiFact_clean_dirty/           # loose files: sample subset only
        ├── 000441.jpg
        └── ...
```

### Splits

| File | Role | Rows (approx.) | Images (approx.) |
|------|------|----------------|------------------|
| `ArtiFact_clean.csv` | Clean / ground-truth pool | ~391k | ~386k |
| `ArtiFact_clean_dirty.csv` | Benchmark with injected errors | ~260k | ~293k |

**`ArtiFact_clean`** holds normalized metadata without deliberate corruption. Use it as a training pool, reference data, or source of ground-truth labels.

**`ArtiFact_clean_dirty`** is the evaluation split. Roughly half of eligible rows carry **one injected error**; the other half are **clean control rows** with no injection. See [How the dirty file works](#how-the-artifact_clean_dirty-file-works) below.

### How the `ArtiFact_clean_dirty` file works

Each row in `ArtiFact_clean_dirty.csv` contains **both** the correct metadata and (when applicable) the corrupted version used for benchmarking.

| What you need | Where to look | Meaning |
|---------------|---------------|---------|
| **Correct / ground-truth value** | Base column (e.g. `location`, `artist_name`, `date_begin`) | The true catalog value before injection |
| **Corrupted / benchmark value** | Matching `*_error` column (e.g. `location_error`, `artist_name_error`) | The value with the synthetic error applied |
| **Error category** | `error_type` | High-level field group (e.g. `place_error`, `artist_error`, `image_error`) |
| **Specific injection** | `error_subtype` | How the error was created (e.g. `city_level_swap`, `century_shift`) |
| **Clean row (no error)** | `error_type` and `error_subtype` are **empty** | All `*_error` columns are empty; base columns hold the true values |

**Example — row with a place error:**

| Column | Value |
|--------|-------|
| `object_ID` | `RIJKS_200270633` |
| `error_type` | `place_error` |
| `error_subtype` | `city_level_swap` |
| `location` *(correct)* | `northern netherlands` |
| `location_error` *(corrupted)* | `delft` |

**Example — clean control row:**

| Column | Value |
|--------|-------|
| `object_ID` | `AIC_180757` |
| `error_type` | *(empty)* |
| `error_subtype` | *(empty)* |
| `location` | `france` |
| `location_error` | *(empty)* |

For **image swaps**, the correct image URL is in `image_url` and the swapped URL is in `image_url_error`. The on-disk file for the corrupted image may use an `-e` suffix (e.g. `000002-e.jpg`); see `image_object_id_error` for the source object.

**Typical evaluation workflow:**

1. Present the model with the **corrupted** view — read from `*_error` columns when non-empty, otherwise from the base column.
2. Ask whether an error exists and which field is wrong.
3. Score against `error_type` / `error_subtype` and the base `*` columns as ground truth.

```python
import pandas as pd

row = pd.read_csv("data/ArtiFact_clean_dirty.csv").iloc[0]

has_error = bool(str(row["error_type"]).strip())
print("injected error?", has_error)
print("category:", row["error_type"], "subtype:", row["error_subtype"])
print("correct location:", row["location"])
print("corrupted location:", row["location_error"])
```

`ArtiFact_clean.csv` has **no** `error_type`, `error_subtype`, or `*_error` columns — only ground-truth metadata.

### Sample index files

For quick experiments without loading the full dirty split, we provide **row-index samples** of `ArtiFact_clean_dirty.csv`:

| File | Rows | Description |
|------|------|-------------|
| `ArtiFact_clean_dirty_sample_10.csv` | 200 | 10 rows per `(error_type, error_subtype)` pair + 10 clean rows |
| `ArtiFact_clean_dirty_sample_200.csv` | 4,000 | 200 rows per pair + 200 clean rows |
| `ArtiFact_clean_dirty_sample_500.csv` | 10,000 | 500 rows per pair + 500 clean rows |

Each sample file is a single-column CSV of **`row_id`** values: the **1-based row index** into `ArtiFact_clean_dirty.csv` (header excluded). Rows are drawn uniformly at random within each error category, restricted to records that have a **local image on disk**.

The corresponding images are available as **loose files** under `data/images/ArtiFact_clean_dirty/` on the Hub (see [Images](#images)). You do not need to download the full `ArtiFact_clean_dirty.tar` to run sample benchmarks.

To materialize a sample as a full metadata table:

```python
import pandas as pd

dirty = pd.read_csv("data/ArtiFact_clean_dirty.csv")
sample_ids = pd.read_csv("data/ArtiFact_clean_dirty_sample_500.csv")["row_id"]
subset = dirty.iloc[sample_ids - 1].reset_index(drop=True)
```

You can generate new samples with the project script `scripts/sample_benchmark_by_error.py` (see `11_sample_benchmark_by_error.sh`).

### Source museums (by `object_ID` prefix)

| Prefix | Institution |
|--------|-------------|
| `RIJKS_` | Rijksmuseum |
| `MET_` | Metropolitan Museum of Art |
| `AIC_` | Art Institute of Chicago |

## Images

Images are provided in **two forms**:

| Format | Location | Contents |
|--------|----------|----------|
| **Loose files** | `data/images/ArtiFact_clean_dirty/` | Images for the **sample index files** only (~200–10k files, depending on sample size) |
| **Tar archives** | `data/images/ArtiFact_clean.tar`, `data/images/ArtiFact_clean_dirty.tar` | **Complete** image sets for each split (~386k / ~293k files) |

Loose files let you run quick experiments on `ArtiFact_clean_dirty_sample_*.csv` without downloading the full archives. For the full CSVs or the clean split, extract the corresponding tar.

### Quick start (samples only)

If you only need a sample benchmark, download the metadata CSVs, a sample index file, and the loose images under `data/images/ArtiFact_clean_dirty/`:

```bash
huggingface-cli download deem-data/ArtiFact \
  --repo-type dataset \
  --include "data/ArtiFact_clean_dirty.csv" \
  --include "data/ArtiFact_clean_dirty_sample_200.csv" \
  --include "data/images/ArtiFact_clean_dirty/*"
```

Then resolve images via `image_path` as usual (paths point at `data/images/ArtiFact_clean_dirty/NNNNNN.jpg`).

We provide three evaluation samples (2000, 4000, and 10000 records) with balanced clean and error-injected records across all 19 subtypes (100, 200, and 500 per subtype, respectively); our baseline uses the 4000-record sample.

### Full dataset (tar archives)

Extract both archives so that paths match the `image_path` column in the full CSVs:

```bash
mkdir -p data/images
tar -xf data/images/ArtiFact_clean.tar -C data/images/
tar -xf data/images/ArtiFact_clean_dirty.tar -C data/images/
```

This yields:

```
data/images/ArtiFact_clean/000001.jpg
data/images/ArtiFact_clean_dirty/000001.jpg
```

`ArtiFact_clean` contains only clean records, while `ArtiFact_clean_dirty` contains an equal split of clean and erroneous records with ground truth labels.

Image files are named by row index (`NNNNNN.jpg`). Rows with image-swap errors may use a companion file such as `000002-e.jpg`.

### Download with the Hugging Face CLI

```bash
pip install -U huggingface_hub
huggingface-cli download deem-data/ArtiFact --repo-type dataset --local-dir ArtiFact
cd ArtiFact
mkdir -p data/images
tar -xf data/images/ArtiFact_clean.tar -C data/images/
tar -xf data/images/ArtiFact_clean_dirty.tar -C data/images/
```

## Data fields

### Shared metadata columns

| Column | Description |
|--------|-------------|
| `object_ID` | Unique record identifier (`RIJKS_…`, `MET_…`, `AIC_…`) |
| `title` | Object title |
| `object_name` | Object type / classification |
| `date_begin`, `date_end` | Creation date range (year) |
| `date_begin_bce`, `date_end_bce` | BCE flags for date fields |
| `materials`, `techniques` | JSON-like list fields |
| `dimensions_json` | Parsed dimensions |
| `culture`, `location` | Cultural and geographic attribution |
| `artist_name`, `artist_role`, `artist_nationality` | Artist metadata |
| `artist_date_begin`, `artist_date_end` | Artist life dates |
| `subjects` | Subject keywords |
| `description`, `inscriptions` | Textual description and inscriptions |
| `image_url` | Original museum image URL |
| `image_path` | Repo-relative path to the local image file |

### Additional columns in `ArtiFact_clean_dirty.csv`

These columns exist **only** in the dirty split where 50% rows have an injected error. See [How the dirty file works](#how-the-artifact_clean_dirty-file-works) for how base vs `*_error` columns relate.

| Column | Description |
|--------|-------------|
| `error_type` | Injected error category (`date_error`, `artist_error`, `culture_error`, `place_error`, `image_error`, …). Empty on clean rows. |
| `error_subtype` | Specific injection recipe (e.g. `city_level_swap`, `century_shift`). Empty on clean rows. |
| `*_error` | Corrupted value paired with the base column of the same name. Empty when that field was not corrupted. |
| `image_object_id_error` | `object_ID` of the artwork whose image was swapped in, when `error_type` is `image_error`. |

**Base column ↔ error column pairs** (correct value on the left, corrupted on the right):

| Correct (ground truth) | Corrupted (erroneous input) |
|--------------------------|---------------------------|
| `date_begin`, `date_end`, `date_begin_bce`, `date_end_bce` | `date_begin_error`, `date_end_error`, … |
| `artist_name`, `artist_role`, `artist_nationality`, … | `artist_name_error`, `artist_role_error`, … |
| `materials`, `techniques` | `materials_error`, `techniques_error` |
| `dimensions_json` | `dimensions_json_error` |
| `culture`, `location` | `culture_error`, `location_error` |
| `image_url` | `image_url_error` |
| `title`, `description` | `title_error`, `description_error` *(propagated text; secondary)* |

### Injected error subtypes

Errors are drawn from a fixed distribution over metadata and image fields, including:

- **Dates:** `century_shift`
- **Dimensions:** `scale_error`, `aspect_swap`
- **Materials:** `material_anachronism`, `material_interchange`
- **Techniques:** `technique_anachronism`, `technique_interchange`
- **Artists:** `artist_tier_1_easy` … `artist_tier_4_hardest`
- **Culture / place:** `culture_tight_swap`, `culture_continent_swap`, `country_level_swap`, `city_level_swap`
- **Images:** `image_tier_1_easy`, `image_tier_2_medium`, `image_tier_3_hard`, `embedding_swap`

## Usage example

```python
import pandas as pd
from pathlib import Path

root = Path("ArtiFact")
clean = pd.read_csv(root / "data" / "ArtiFact_clean.csv")
dirty = pd.read_csv(root / "data" / "ArtiFact_clean_dirty.csv")

# Resolve an on-disk image from image_path
sample = dirty.iloc[0]
img = root / sample["image_path"]
print(sample["object_ID"], sample["error_type"], sample["error_subtype"], img.exists())
```

## Notes

- **Museum terms:** Underlying images and metadata originate from public museum APIs and web collections. Users must comply with each institution's terms of use and attribution requirements when redistributing or publishing results.
- **Archive layout:** For full evaluation, images must be extracted from the `.tar` files before all `image_path` values resolve on disk.
- **Error injection:** The dirty split contains synthetic errors for benchmarking; it does not reflect real cataloging mistakes at the source museums.
- **Images:** Not every row has a local image; check `image_path` and file existence before image-based evaluation.

## Citation

If you use ArtiFact in your work, please cite the dataset and acknowledge the source museums (Rijksmuseum, Metropolitan Museum of Art, Art Institute of Chicago).

```bibtex
@article{duarte2026artifact,
  title     = {{ArtiFact}: A Large-Scale Multi-Modal Cultural Heritage Dataset},
  author    = {Duarte, Luciano and Ovcharenko, Olga and Schelter, Sebastian},
  year      = {2026},
  url       = {https://github.com/OlgaOvcharenko/ArtiFact}
}
```

## Contact

Repository: [deem-data/ArtiFact](https://huggingface.co/datasets/deem-data/ArtiFact)
Website: [https://olgaovcharenko.github.io/ArtiFact/](https://olgaovcharenko.github.io/ArtiFact/)