Unified Neuromorphic Datasets Repository (UNDR) is a tool that makes it easy to download and analyse a large number of Neuromorphic datasets.
Specifically, UNDR:
- defines a standard format to store Neuromorphic datasets.
- re-destributes as many datasets as possible, encoded in the aforementioned format.
- provides a graphical app, command-line tools, and libraries (Python and Rust) to download the datasets.
- provides a Python library to implement parallel processing algorithms.
- Browse datasets online: https://www.undr.space
- Download the app: https://github.com/neuromorphicsystems/undr/releases
- Python documentation: https://neuromorphicsystems.github.io/undr/
- Rust documentation: https://docs.rs/undr/0.2.0/undr/
- Self-hosting guide: https://neuromorphicsystems.github.io/undr/
- AWS S3 hosting guide: https://neuromorphicsystems.github.io/undr/
- Converting a dataset to UNDR: https://github.com/neuromorphicsystems/undrdg
- UNDR format specification: https://neuromorphicsystems.github.io/undr/
See app/README.md for build instructions.
See aws/README.md for build instructions.
python3 -m venv .venv
source .venv/bin/activate
python documentation/build.pyThe documentation is generated in documentation/_build and the entry point is documentation/_build/index.html.
See python/README.md for build instructions.
See rust/README.md for build instructions.
The Python scripts in the top-level scripts directory extract version numbers to generate GitHub releases. Python scripts that download or analyse datasets can be found in python/examples.
The top-level specification directory contains the JSON schemas at the heart of the UNDR format specification. The full specification includes a few rules that are not enforced by the schemas (due to limitations of JSON schema implementations at the time of writing) and a description of binary encodings. See https://neuromorphicsystems.github.io/undr/ for details.