🚀 Launched open source active learning toolkit for computer vision
Hi Indie Hackers! I am Nikolaj from Encord and I want to share a cool project I have been working on for the last year that I am super excited about 👏
Working with many leading AI companies over the past 2 years, we found that many of their AI models performed poorly and unpredictably in the real world, and they lacked the tools to understand why 🤔
To help them better understand why and mitigate it we built our new open source active learning toolkit for computer vision, Encord Active.
Encord Active is an active learning toolkit that helps you diagnose and improve your data quality (outliers, duplicates, data imbalance), label quality (erroneous labels, edge-cases, class imbalance), and model quality (model failure modes, edge-cases, performance analysis).
Some of the things you can do with the tool:
- Find hidden model failure modes 🔎
- Find and fix data & label errors ⚠️
- Curate high-impact data from your data pool 🔖
- Balance your dataset ⚖️
It is still very much a work in progress, but we want it out in the world so that it can help as many AI companies as possible deploy more robust models to production.
If you want to support the project we would love a ⭐on GitHub– this mean a lot to us on its own.
"Can you show me the product?" - Sure Check out this video.
Cheers,
