Open source active learning framework to improve model performance
Encord Active is an open-source active learning framework used to create actionable workflows to improve data quality with real-world data and labels.
Use Encord Active to find failure modes in your models, prioritize high-value data for re-labeling, and drive smart data curation to improve your model performance. The toolkit is flexible, extensible, and scalable.
- Discover errors and outliers within your data and labels.
- Evaluate your model performance by 25+ quality metrics (annotation quality, brightness, object size, etc.) to discover which features matters for your model.
- Create your own custom quality metrics.
- Prioritize different actions to improve your dataset (re-labeling, augmenting, or gathering data).
Find the repo here:
https://github.com/encord-team/encord-active
