Using the MegaDetector Pipeline on AML

Scheduled Pipeline Execution

The primary usage of the MegaDetector pipeline is through a scheduled endpoint, which triggers automatically every Monday and Thursday. This scheduled deployment ensures that the pipeline processes incoming data consistently without manual intervention.

Manual Pipeline Execution

While the pipeline typically runs automatically, manual execution is possible:

  1. Trigger the Pipeline: Go to the Pipelines section in AML Studio.
  2. Run the Pipeline: Select the pipeline endpoint and click Submit

Pipeline Phases

  • Phase 1: Performs object detection using the cameratraps_pipeline.py script.
  • Phases 2 & 3: Handles post-processing and archival.

Accessing Results

  • After execution, the processed results will be available in the bronze_megadetector datastore.
  • You can access detailed logs and output files directly from AML Studio.