The Quick, Draw! Dataset
A public dataset of 50 million vector drawings across 345 categories, captured from the Quick, Draw! game.
What is The Quick, Draw! Dataset?
The Quick, Draw! Dataset is a public collection of 50 million vector drawings gathered from the Quick, Draw! game, where players were prompted to sketch simple objects. It provides timestamped stroke data with metadata like country and recognition status, serving as a valuable resource for training drawing recognition models and analyzing human sketching patterns. The dataset addresses the need for large-scale, real-world drawing data in machine learning and creative computing research.
Machine learning researchers, data scientists, and developers working on sketch recognition, generative AI, or creative coding projects. It's also used by artists and educators for data visualization and interactive installations.
Developers choose this dataset because it offers a massive, curated collection of real human drawings with rich temporal and geographic metadata, available in multiple preprocessed formats for immediate use. Its open CC BY 4.0 license and Google Cloud Storage hosting make it easily accessible for both academic and commercial projects.
Overview
Documentation on how to access and use the Quick, Draw! Dataset.
Use Cases
Best For
- Training neural networks for sketch recognition and classification
- Researching human drawing behavior and cultural differences in sketching
- Building creative AI applications that generate or interact with drawings
- Educational projects teaching machine learning with real-world datasets
- Data visualization projects exploring large-scale human-generated content
- Developing tools for sketch-based search or retrieval systems
Not Ideal For
- Projects requiring perfectly clean and curated data for sensitive applications, due to the dataset's warning about potential inappropriate content.
- Real-time or interactive systems needing low-latency data streaming, as the dataset is stored in large batch files on cloud storage.
- Applications that depend on color or 3D sketch data, since the dataset exclusively contains 2D vector drawings without color information.
Pros & Cons
Pros
With 50 million drawings across 345 everyday object categories, it offers a vast resource for training robust models, as detailed in the dataset's description and categories.txt file.
Each drawing includes timestamp, country code, and recognition status, enabling research on cultural and temporal patterns in sketching, per the raw data format specification in the README.
Available in raw NDJSON, simplified vectors, binary files, and pre-rendered numpy bitmaps, providing multiple entry points for different use cases, as listed in the preprocessed dataset section.
Cons
The README explicitly states that while moderated, the dataset may contain inappropriate content, which can be a limitation for family-friendly or professional applications requiring high data purity.
Accessing and processing the data requires familiarity with Google Cloud Storage and command-line tools like gsutil, and managing the large file sizes can be resource-intensive, as noted in the download instructions.
Frequently Asked Questions
Found a gem we're missing?
Open-Awesome is built by the community, for the community. Submit a project, suggest an awesome list, or help improve the catalog on GitHub.