BIIGLE

BIIGLE provides image and video annotation for marine science to generate curated annotations and training data for observational analyses and machine-learning applications.

Key Features:

• Manual and computer-assisted annotation: Supports manual full-frame image and video annotation alongside computer-assisted techniques to annotate frames and collect training data for machine-learning-based (semi-)automatic annotation.
• Video annotation: Includes capabilities for annotating video sequences and individual frames extracted from video.
• Large-image and large-scale dataset support: Handles large gigapixel images and large-scale imaging datasets spanning extensive areas and long-term observation periods.
• Collaborative taxonomies and label tree collaboration: Enables custom taxonomies, enhanced label tree collaboration, and quality-control workflows to maintain consistent annotations across users and projects.
• Scalability and deployment: Can be deployed in cloud platforms, local networks, or on mobile devices to support diverse processing and annotation environments.
• Machine learning-assisted annotation: Integrates machine learning-assisted image annotation to increase annotation throughput and support automated workflows.
• Application instance federation: Supports application instance federation to integrate annotation instances across different research groups and projects.

Scientific Applications:

• Large-scale marine imaging analysis: Facilitates annotation and analysis of extensive marine imaging datasets for observational studies and long-term monitoring.
• Training data generation for machine learning: Produces curated annotations and labeled datasets for developing and validating machine-learning and computer-vision methods.
• Consistent multi-user annotation: Supports collaborative taxonomy development and quality control to enable reproducible comparative studies across teams.

Methodology:

Combines manual full-frame image and video annotation with computer-assisted techniques to collect training data for machine-learning-based (semi-)automatic annotation.