Your Goals with Encord

Before doing anything with Encord you should create goals of what exactly you intend to accomplish with the platform. Your goals inform how best to use Encord.

​

Data Curation with Index

Data curation involves selecting the best data units for use in other down stream activities. Encord helps you with data selection in a number of ways. First, using Encord Index you can visually select and group your data. But this can be time consuming if the amount of data you are trying to group gets over a few thousand data units. This is where native and custom metadata come into play. In Index and Active, metadata helps to filter and sort your data so you can narrow your data search quickly. Native metadata is metadata that is typically included or embedded with your data unit. For example, DICOM series include embedded metadata that appears while annotating the series. Custom metadata is metadata added to the data unit during or after data imports.

Before importing your data think about how you want to search for and filter your data. Custom metadata can always be added to your data after your initial data import, but starting with a strong foundation of custom metadata (importing with your data) can significantly increase your the time to ROI and achieving your goals.

​

Annotation with Annotate

Annotation depends on your selected data (Datasets) and how you want that data annotated. Encord Annotate supports annotating a number of data modalities (images, image groups and sequences, videos, DICOM series, audio files, and text) and methods of annotation.

Workflows specify who performs annotations and reviews and the process of the annotating and reviewing data.

Encord Annotate provides fast and accurate automated annotation using SAM 2 auto segmentation and tracking. Using Editor Agents you can train and use models to perform automated annotation.

Task Agents provide a mechanism to bring your own models into Workflows for annotation and reviews.

​

Label Validation and Model Optimization with Active

After annotating your data units in Annotate, use Active to perform label error detection and validation on the labels. Active provides a number of label, data, and model quality metircs, anlaytics, and custom filtering criteria support for label error detection and validation.

After training your model on labels exported from Annotate, import and compare your model predictions in Encord Active. Discover where your models are weak and improve their performance using analytics, quality metrics, and embedding plots. As in Index, custom filtering criteria helps you quickly search and select data that can improve your model’s performance.