OSOceanAcoustics / echodataflow

Orchestrated sonar data processing workflow
https://echodataflow.readthedocs.io/en/latest/
MIT License
4 stars 1 forks source link

Provide an option to read 2 datastores with different purposes. #125

Open valentina-s opened 3 weeks ago

valentina-s commented 3 weeks ago

Description

Currently the pipeline.yaml configuration provides an option to read two paths with the datastore option. However, if two paths are passed the code goes into the steps needed to parse the two frequencies of Lasker, and that is a very specific scenario. In the context of echoshader_flow_predictions, it needs to also process

Expected Behavior

Ideally we should be able to pass several paths and decide what to do with them within the stages. As a starting point we should at least be able to provide an inidicator whether we are dealing with two paths because we want to combine channels for MVBS or we want to combine MVBS with masks.

Actual Behavior

On realtime branch we call a function get_input_from_store_folder. This function checks if there are more than one folders and combines the frequencies so that it works for Lasker. On echoshader_predictions branch we call a function get_input_from_store_folder_mvbs_score which is combining mvbs and score for echoshade_flow_predictions flow.

The above steps are happening in the initialization_flow.

Possible Solution / Suggestion

A short-term solution is to use keys to indicate if the paths are for MVBS frequency, or score combination.

To run echoshader on Lasker we need to access 3 folders: two MVBS frequency folders and Score folders.