The open sourceactive learning toolkit for computer vision

A customizable toolkit for your data, labels, and models. Encord Active helps you find failure modes in your models, prioritize high-value data for labelling, and drive smart data curation to improve model performance.

GitHub  GithubRead the docsJoin our Discord  Discord
  • 🚀The fastest way to debug your dataset and improve model performance
  • 🤠Built for ML Engineers and Data Scientists
  • One line of code to get started to import you data, labels, and model predictions
$ pip install encord-active
$ encord-active quickstart
How it works

Enhance data quality and model performance

Encord Active has been designed as a all-in-one open source toolkit for improving your data quality and model performance. Use the intuitive UI to explore your data or access all the functionalities programmatically.

Data Quality

Ensure high quality training data for computer vision

Discover errors, outliers, and edge-cases within your data - all in one open source toolkit.

Uses platform

Understand your data distribution

Get a high level overview of your data distribution, explore it by customizable quality metrics, and discover any anomalies.

Uses platform

Surface outliers

Quickly find image outliers and tag them for further investigation.

Uses platform

Find similar images

Use powerful similarity search to find more examples of edge-cases or outliers.

Label Quality

Ensure your annotations meet your quality standards

Save time and improve label quality by understanding your label distribution, label errors, and annotator performance.

Png feature model import

Understand your label distribution

Explore your label distribution by label classes, quality metrics, and custom metrics to uncover hidden gaps in your data.

Downloading models

Find outliers in your labels

Quickly find label outliers and tag them for further investigation.

Downloading models

Explore annotator performance

Analyze the speed and performance of your annotators.

Model Quality

Gain visibility into model performance and failure modes

Benefit from intelligent model evaluation features based on quality metrics that allows you to find model failure modes and target your data collection and labeling on high-value data.

Quality prediction iou grid in program

Find model failure modes

Drill down into each quality metrics to discover how your model is performing and find hidden model failure modes.

Png feature model import

Correlate model performance

Correlate model performance to each quality metric to understand what data and label characteristics impact your model performance.

Png review json

Find & fix label errors

Find label errors and send them to re-labeling immediately to improve the quality of your training data.

Any questions?

Join our Discord to get in touch with the team