Hafftka/michael-hafftka-catalog-raisonne

Michael Hafftka – Catalog Raisonné Dataset (1970s–2025)

A rare, large-scale dataset of ~3,800 paintings by a single artist, spanning over five decades.

This dataset provides a longitudinal view of artistic development, combining images with structured metadata (title, year, medium, dimensions, collection) and including works held in major museum collections.

Example from the dataset (1995)

Why this dataset is unique

• Single-artist consistency: Unlike most art datasets, all works are by one artist
• Long time span: Covers work from the 1970s to the present
• Museum-provenance works included: Pieces held by major institutions
• Structured metadata: Suitable for search, filtering, and modeling
• Multiple image views: Includes details and alternate perspectives

Potential use cases

• Training LoRA or style models
• Studying artistic evolution over time (1970s–2025)
• Building image + metadata retrieval systems (RAG)
• Research in digital humanities or art history
• Experiments in generative art and style transfer

These use cases are enabled by the combination of images and structured metadata for each work.

Dataset Summary

This dataset contains approximately 3,780 digitized artworks from the archive of Michael Hafftka, forming part of an ongoing catalog raisonné.

The collection spans over 50 years of work and includes both widely exhibited and lesser-known pieces, with structured metadata enabling analysis, retrieval, and machine learning applications.

Artist

Michael Hafftka is an American painter whose work spans several decades, beginning in the 1970s. More than 100 works are held in major museum collections, including:

• The Metropolitan Museum of Art
• The Museum of Modern Art (MoMA)
• SFMOMA
• The British Museum

Metadata Fields

Field	Description	Example / Notes
catalog_number	Unique archive identifier	e.g., ####, ####.##, p###, p###.##
file_name	Relative path to the artwork image file	e.g., images/<filename>
artist	Fixed: "Michael Hafftka"
title	Artwork title
year	Creation year (exact when known; "cYYYY" for the rare approximate/circa cases e.g. c1990; blank if unknown)	1975, c1990
medium	Materials/technique	e.g., "oil on linen", "acrylic on canvas"
dimensions	Physical dimensions of the artwork (in inches or cm)	e.g., "78in × 62in"
collection	Holding institution (if public/museum)	e.g., "Metropolitan Museum of Art"
copyright_holder	Typically "Michael Hafftka"
license	Dataset/image license	"CC-BY-NC-4.0"
view	Specific view photographed	e.g., "front", "detail", "left panel"

License

This dataset is released under the CC-BY-NC-4.0 license.

Images may be used for research and other non-commercial purposes with proper attribution.

Source

Images and metadata are drawn from the personal archive of artist Michael Hafftka.

Notes

Metadata completeness varies across the dataset. Earlier works are more likely to have missing fields such as medium and dimensions, reflecting the realities of long-term archival documentation.

This variability may be useful for:

• studying incomplete or evolving archives
• training models to handle missing metadata
• research in historical documentation practices

Image quality also varies, as the dataset includes documentation produced over several decades using different photographic and scanning methods.

Time Coverage

The artworks represented in this dataset span from the 1970s through 2025.

How to Load and Use (Python Example)

Install the library if needed:

pip install datasets matplotlib

Then run this code in Python (e.g. in Google Colab – go to colab.research.google.com, make a new notebook, paste into a cell, press Shift+Enter):

from datasets import load_dataset
import matplotlib.pyplot as plt

# Load the dataset
ds = load_dataset("Hafftka/michael-hafftka-catalog-raisonne")

# Print basic info
print(ds)
print("Number of images:", len(ds["train"]))

# Display one image with its title and year
example = ds["train"][0]  # change 0 to any number to see different artworks
plt.imshow(example["image"])
plt.title(f"{example['title']} ({example.get('year', 'Unknown')})")
plt.axis("off")
plt.show()

# Example: filter works from the 1980s
filtered = [x for x in ds["train"] if x["year"] and "198" in x["year"]]

print("Number of works from 1980s:", len(filtered))

# Example: find all works in museum collections
museum_works = [x for x in ds["train"] if x["collection"]]

print("Number of works in collections:", len(museum_works))

# Show one example
example = museum_works[0]
plt.imshow(example["image"])
plt.title(f"{example['title']} – {example['collection']}")
plt.axis("off")
plt.show()

# Example: group works by decade
from collections import Counter

def get_decade(year):
    if not year or not year.isdigit():
        return "Unknown"
    return year[:3] + "0s"

decades = Counter(get_decade(x["year"]) for x in ds["train"])

print(decades)