What is Dioptra?
Dioptra is a versatile, open-source platform for data curation and management, specifically crafted for computer vision, natural language processing (NLP), and large language models (LLMs). It empowers users to curate essential unlabeled data, manage metadata, diagnose model performance issues, and seamlessly integrate with labeling and retraining processes.
How to use Dioptra?
1. Curate the most critical unlabeled data to enhance model accuracy and domain coverage.
2. Register your metadata with Dioptra to keep your data secure and accessible.
3. Diagnose and understand model failure modes and performance regressions using Dioptra's diagnostic tools.
4. Utilize active learning miners to select the most valuable unlabeled data.
5. Integrate Dioptra with your labeling and retraining workflows using its comprehensive APIs.