Post-Processing and Data Transfer Documentation

Overview

This documentation provides detailed information on the post-processing and data transfer stages for the MegaDetector and BirdNet pipelines. These stages ensure that detected objects (for MegaDetector) and segmented audio detections (for BirdNet) are filtered, organized, and stored in final Azure Data Lake Storage containers for easy access, retrieval, and analysis.

Documentation Contents

• MegaDetector Post-Processing: Details on the final steps in the MegaDetector pipeline, including arguments, configurable parameters, CSV output structure, and recommendations for changes to confidence thresholds and archiving options.
• BirdNet Post-Processing: Information on the BirdNet segmentation step, covering key arguments, parameter choices, CSV output columns, and how to adjust settings to optimize data storage and quality.

Key Topics

• Final Containers: Explanation of the storage containers used for processed and archived data in each pipeline.
• Arguments and Customization: Insights into configurable parameters, including confidence thresholds, directory prefixes, and batch naming for easy identification.
• CSV Output Structure: A breakdown of each CSV output format, providing details on columns such as confidence, bounding_box, species_id, and timestamps.

These files serve as a comprehensive guide for understanding, modifying, and managing the final processing stages within each pipeline.