Labelbox offers several ways to import annotations and attach them to a dataset in Labelbox.
To speed up your labeling efforts, you can use the Model-assisted labeling (MAL) workflow to import your model predictions as pre-labels. As soon as you create an ontology and upload a dataset, you can import annotations via the MAL workflow. Each annotation you import will need to reference an annotation class and a Data Row in Labelbox.
If you are switching to Labelbox and need to migrate your human-made ground truth annotation from another platform, you can use the Import ground truth method.
If you want to assess your model performance, you can use Model-assisted labeling and Ground truth import to import your model predictions and ground truth annotations into Labelbox. Then, use Model diagnostics to understand where your model performs well and where it needs more training.
We recommend importing any external annotations via Python Annotation types. Python annotation types provide a seamless transition between local modeling and Labelbox. Some of the helper functions include:
- Build annotations locally with local file paths, numpy arrays, or urls and create data rows with a single line of code
- Easily upload model predictions for Model-assisted labeling or Model Diagnostics by converting annotation objects to the NDJSON import format
- Helper functions for drawing annotations, converting them into shapely objects
You can also import your annotations by formatting your annotations in an NDJSON file.
Updated 8 days ago