handwriting-test-v2

This dataset contains handwriting stroke data collected using a stylus (S Pen) on a tablet device. Optimized for training RNNs (Recurrent Neural Networks) on handwriting generation/recognition tasks.

Data Format

Each row in the Parquet files represents a complete handwriting sample:

Column	Type	Description
id	string	Unique identifier (UUID)
text	string	The prompt text that was written
dx	string (JSON array)	Delta X offsets between consecutive points
dy	string (JSON array)	Delta Y offsets between consecutive points
eos	string (JSON array)	End-of-stroke flags (1 = pen lift, 0 = continue)
scale	double	Scale factor used for normalization
created_at	string	ISO timestamp of creation
session_id	string	Collection session identifier

RNN Training Format

The stroke data is stored in the format commonly used for RNN handwriting models:

• dx/dy: Position deltas from the previous point (first point has dx=dy=0)
• eos: Binary flag indicating pen lifts (end of stroke)
• Data is normalized by bounding box for consistent scale

Visualization

Preview SVGs are available in renders_preview/ for HuggingFace Dataset Viewer.

Usage

from datasets import load_dataset
import json

# For private repos, use: load_dataset("finnbusse/handwriting-test-v2", token="YOUR_HF_TOKEN")
dataset = load_dataset("finnbusse/handwriting-test-v2")

# Access a sample
sample = dataset['train'][0]

# Parse stroke data
dx = json.loads(sample['dx'])
dy = json.loads(sample['dy'])
eos = json.loads(sample['eos'])

# Reconstruct absolute positions
x, y = 0, 0
positions = []
for dx_i, dy_i, eos_i in zip(dx, dy, eos):
    x += dx_i
    y += dy_i
    positions.append((x, y, eos_i))

Collection Method

Data was collected using a web application with Pointer Events API, capturing stylus input including pressure and tilt when available.