availai
availai
is a Python library for managing and building computer vision workflows, including tools for dataset management, preprocessing, and seamless integration with Weights & Biases (W&B) and Roboflow. Future updates will expand functionality to include broader machine learning, deep learning, audio processing, NLP, and large language models (LLMs).
Features
Dataset Management
-
Download Datasets:
- Fetch datasets directly from Weights & Biases using artifact versioning.
- Retrieve datasets from Roboflow in various formats (e.g., YOLOv8).
-
Upload and Log:
- Upload datasets as W&B artifacts for versioning and tracking.
- Log dataset-related tables for visualization in W&B projects.
-
Dataset Customization:
- Create smaller dataset versions for quick testing by reducing one or more dataset splits (e.g.,
train
, val
, test
).
-
Copy Datasets:
- Copy datasets to a specified location with a progress bar for visibility.
Integration with Roboflow
- Download datasets using workspace and project names.
- Support for multiple dataset formats, such as YOLOv8, COCO, and more.
Integration with Weights & Biases
- Manage datasets using W&B artifact tracking.
- Log preprocessing steps and metadata as W&B artifacts for reproducibility.
Installation
Install availai
from PyPI:
pip install availai