MetFaces Dataset

MetFaces is an image dataset of human faces extracted from works of art.

The dataset consists of 1336 high-quality PNG images at 1024×1024 resolution. The images were downloaded via the Metropolitan Museum of Art Collection API, and automatically aligned and cropped using dlib. Various automatic filters were used to prune the set.


All data is hosted on Google Drive:

Path Size Files Format Description
metfaces-dataset 14.5 GB 2621 Main folder
├ metfaces.json 1.8 MB 1 JSON Image metadata including original download URL.
├ images 1.6 GB 1336 PNG Aligned and cropped images at 1024×1024
└ unprocessed 13 GB 1284 PNG Original images

Reproducing the dataset

MetFaces 1024x1024 images can be reproduced with the script as follows:

  1. Download the contents of the metfaces-dataset Google Drive folder. Retain the original folder structure (e.g., you should have local/path/metfaces.json, local/path/unprocessed.)
  2. Run --json data/metfaces.json --source-images data --output-dir out


The metfaces.json file contains the following information for each image:

    "obj_id": "11713",                                  # Metmuseum object ID
    "meta_url": "",
    "source_url": "",
    "source_path": "unprocessed/image-11713.png",       # Original raw image file under local dataset copy
    "source_md5": "c1e4c5a42de6a4d6909d3820c16f9eb5",   # MD5 checksum of the raw image file
    "image_path": "images/11713-00.png",                # Processed 1024x1024 image
    "image_md5": "605a90ab744bdbc9737da5620f2777ab",    # MD5 checksum of the processed image
    "title": "Portrait of a Gentleman",                 # Metmuseum object's title
    "artist_display_name": "Charles Willson Peale",     # Metmuseum object's artist's display name
    "face_spec": {                                      # Info about the raw image:
      "rect": [404, 238, 775, 610],                     # - Axis-aligned rectangle of the face region
      "landmarks": [...],                               # - 68 face landmarks reported by dlib
      "shrink": 2
    "face_idx": 0

For full Metmuseum metadata, you can access the meta_url contents by e.g., curl