Introduction to Encord

Encord is a comprehensive data engine and software stack solution offering a range of features to streamline and optimize your ML workflow. It consists of three main components: Annotate, Active, and Index.

Index: Index provides Encord’s storage for your images, image groups, image sequences, videos, and DICOM. Once your data imports into Encord (locally or from cloud storage) you can create Datasets for your Projects based on your data. Your data can belong to multiple Datasets and Projects. Learn more about Index here.

Annotate: Annotate is a powerful tool that enables users to create labeling workflows and transform unlabeled data into labeled data. Label data with object boundaries and classification labels, and review labels using manual or automatic quality assurance methods. Annotate simplifies the process of creating high-quality labeled datasets, which are essential for training ML models. Learn more about Annotate here.

Active: Active facilitates the validation and debugging of datasets through systematic active-learning cycles. Analyze your annotated data sets to hone in on the best labels, while rejecting those hindering a model’s performance. By iteratively refining your training data, Active helps improve the quality and performance of ML models over time. Learn more about Active here.

Encord promotes an iterative approach towards developing optimal computer vision models. By leveraging the capabilities of Annotate, Active, and Index, you can continuously refine your models and labels.

As changes to data are made, the impact on the model’s performance is tracked, enabling you to assess the effectiveness of your modifications. Incorporating these insights results in improved models and more accurate labeling. This feedback loop allows you to build the optimal computer vision model for any use-case.