pysolotools 0.0.1

unity computer vision dataset toolchain

The open-source tool for using Unity SOLO Datasets

---

Introduction

pysolo would allow you to the following:

• Parse SOLO datasets generated with Unity Perception.
• Dataset iterables.
• Dataset statistics.
• Clients to access remote datasets
  • Unity Computer Vision Datasets (UCVD)

Pre-Requisites

• Install Anaconda or Miniconda (recommended). Install pre-commit.
• Make sure pip is installed.

Installation

---

To install, run

pip install pysolo --index-url=https://artifactory.prd.it.unity3d.com/artifactory/api/pypi/pypi/simple

** The package lives in the internal PyPi repo for now.

SOLO Dataset

Local Datasets

from pysolotools.consumers import Solo
solo = Solo(data_path="<<dataset-path>>")

Access frames with FrameIterator:

solo.frames()

Cloud Datasets

Getting API Token

• Create a Unity account if you don't at Unity Services
• Navigate to Projects and create a project if you don't have one.
• Navigate to Projects>Service Account in the dashboard to create a new one.
• Generate your API key.
• Save your API key somewhere safe, it will be used later.
• Navigate to Projects>Select the project to use and record the Project ID.
• Navigate to Projects>Orgaization Settings and record the Organization ID.

Downloading datasets from UCVD

from pysolotools.clients import UCVDClient
client = UCVDClient(
    org_id="unity-org-id",
    project_id="unity-project-id",
    sa_key="sa-key",
    api_secret="api-secret"
)

client.download_dataset_archives("<<dataset-id>>")

Sphinx Docs

Github Pages

To generate docs locally:

1. cd docs/
2. make html

If you want to rebuild the rst files, please run: make apidoc. This will generate the rst files based off docstring comments. (Note: This does not update existing rst files).

Additional Resources

Blog Posts and Talks

• Data-centric AI with Unity Computer Vision Datasets blogpost
• Workshop notebook Notebook

Community and Feedback

The Unity Computer Vision demos are open-source and we encourage and welcome contributions. If you wish to contribute, be sure to review our contribution guidelines and code of conduct.

Support

For feature requests, bugs, or other issues, please file a GitHub issue using the provided templates we will investigate as soon as possible.

Newsletter

License

Apache License 2.0