Locate each face with a bounding box and mark visible landmarks. Apply the same convention across viewpoints and capture conditions.
Separate facial regions with masks or polygons when the training task requires labeled areas rather than individual points.
Unitlab is a training data platform. It helps teams prepare labeled datasets; your downstream models and systems perform verification or recognition.
Facial recognition is one use case within identity and biometrics. This page covers the broader data workflow across images, video, and related documents.
Teams can label image and video evidence and annotate document fields or regions with the appropriate modality tools.
Define classes and labeling instructions that match your task. Specify point placement and how annotators should handle hidden or ambiguous features.
Curate examples across relevant lighting, pose, image quality, and occlusion conditions. Document the selection criteria used for each dataset.
Use multimodal workflows where related images, video, and documents need shared project context and consistent labeling rules.
Reviewers inspect the source evidence and apply shared instructions. Consensus and quality gates can support the review process where appropriate.
Yes. Inspect and filter annotated examples, then preserve dataset versions so teams can trace the data used for model development.
Unitlab documentation provides annotation and integration guidance. Follow the API and SDK instructions for your data pipeline and supported export formats.
We use cookies to improve your experience. See our Privacy Policy and Cookie Policy.