search

GeospatialCentroid / Agroforestry

Generic method for detection of non forested trees within Nebraska
0 stars 0 forks source link

Code Review #11

Open dcarver1 opened 7 months ago

dcarver1 commented 7 months ago

Two options of code review

This is a reasonable summary of the workflow at 2024-03-08

Mission : map all the trees in Nebraska in 2010, 2016, 2020 using ~1m NAIP imagery.

Improve a wonky workflow?

image

See flowcharts/flowworkflow.drawio file for details (tab 2024-03-01_workflowUpdate)

The moral of the story is this current effort requires

  1. manually setting parameters in a python config file and running a python script
  2. running a few r script (best part :) )
  3. running a jupyter notebook to visually eval results before exporting to GEE Asset
    • waiting 10-20 hours
  4. evaluating results in GEE
  5. export from GEE web interface (likely more processing methods being added here in the near future )
  6. potentially more image aggregation effort in R

It looks terrible, but there are legitimate reasons for all this platform/language hopping

Still I think there potential for some serious wins in ease of use by aggregating some or all of these steps into a more single source workflow. (RMD) feels like the winner.

Python specific review

Before this project I use to think people arguing about the quality of a specific language or making claims about one being better then the other was a silly practice. While I still understand that logic and I think it's true, it's been very hard for me as a primary R programmer to adapt to the python development experience. So I expect that there are silly gems of poor python logic hidden throughout this code based. While the workflow is producing the expected results I would feel much more confident in the process with a second set of eyes evaluating the execution of the python code. This includes functions in the 'agroforestry' sub folder, 0_developingTraingingdata.py and 1_applyModel.ipynb

dcarver1 commented 7 months ago

@matthewross07

Matt, here are my thoughts on what would be a valuable code review for the Agroforestry project at this time. If either of these efforts are worth pursuing by one of the ROSS lab team I'll make an effort to clean up the repo and improve documentation.