TUBerlin

TUBerlin is a SketchDataset loader for the TU Berlin dataset, where vector sketches are stored in .svg files. The sketches are represented with cubic Bézier curves with four control points. The dataset includes sketches across 250 categories. Each category contains 800 sketches.

Source: datasets/tu_berlin.py

Data Format

Each SVG file includes a series of path. Each path comprises commands M (Move), C (CubicBezier), and L (Line).

Directory Layout

After download and extraction, the dataset is expected under:

<root>/
  TUBerlin/
    .metadata.parquet
    sketches_svg.zip
    svg/
      airplane/
        *.svg
      backpack/
        *.svg
      ...
      zebra/
        *.svg

Code

from sketchkit.datasets import TUBerlin

ds = TUBerlin(cislab_source=True)

# Load a sketch using index
sketch = ds[0]
print(sketch.width, sketch.height)
print(sketch.path_num, sketch.curve_num)

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/TUBerlin/sketches_svg.zip

    • Official host: http://cybertron.cg.tu-berlin.de/eitz/projects/classifysketch/sketches_svg.zip

Metadata Columns

  • id: Global unique identifier across all sketches.

  • sub_id: Identifier within the specific category.

  • category: Drawing category name (e.g., “airplane”, “alarm clock”, etc.).

  • split: “train”, “valid”, or “test”.