SketchXPRIS

SketchXPRIS is a SketchDataset loader for the SketchX-PRIS-Dataset, which contains 20,000 drawings across 25 categories. It was collected by SketchX and PRIS to study universal perceptual grouping. Each drawing is stored as a sequence of strokes in stroke-3 format.

Source: datasets/sketchx_pris.py

Data Format

Each category’s data is stored in a .ndjson file. The sketches are parsed from stroke-3 format arrays (N, 3) containing:

  • dx: delta x-coordinate, relative displacement from the previous point

  • dy: delta y-coordinate, relative displacement from the previous point

  • pen_state: pen state (0 for drawing / pen down, 1 for lifting / pen up)

The loader converts the relative coordinates into absolute coordinates, calculates the canvas size based on the bounding box (adding a 10% margin and ensuring a minimum 100x100 size), and groups continuous drawing segments into paths. Each stroke is represented as a cubic Bézier Curve within a Path.

Directory Layout

After download and extraction, the dataset is expected under:

<root>/
  SketchX-PRIS-Dataset-master/
    Group ID/
    Perceptual Grouping/
      apple.ndjson
      ice_cream.ndjson
      ...
  SketchX-PRIS.zip
  .metadata.parquet

  • All sketch files are stored as .ndjson files categorized by class in the Perceptual Grouping/ directory.

  • The original archive SketchX-PRIS.zip is kept in the root directory for integrity checking.

Code

from sketchkit.datasets import SketchXPRIS

ds = SketchXPRIS(cislab_source=True)

sketch = ds[0]
print(sketch.width, sketch.height)
print(sketch.path_num, sketch.curve_num)

# Filter by category
ice_cream_metadata = ds.items_metadata[(ds.items_metadata["category"] == "ice_cream")]
sketches = [ds[row.id] for _, row in ice_cream_metadata[:5].iterrows()]

Arguments

  • cislab_source: Selects the download source. If True, the dataset is downloaded from the CISLAB CDN mirror; otherwise it is downloaded from official host.

    • CISLAB mirror: https://cislab.hkust-gz.edu.cn/projects/sketchkit/datasets/SketchXPRIS/SketchX-PRIS.zip

    • Official host: https://github.com/KeLi-SketchX/SketchX-PRIS-Dataset/archive/refs/heads/master.zip

Metadata Columns

  • id: Global unique identifier across all sketches.

  • sub_id: Identifier within the specific category.

  • category: Drawing category name (e.g., “ice_cream”, “apple”).

  • split: Split in the dataset (e.g., “train”, “valid”, or “test”).

  • label_names: A string of part label names, separated by ‘;’ (e.g., “part1;part2;part3”).