Skip to content

Task 2: Classification

We will use the OFO tools Geograypher and Tree Classification Framework to generate predictions of tree species and live/dead status for detected trees. We will use all of the views of each detected tree from each drone photo it appears in. Because the raw drone photos are not geospatial data products, some intermediate processing using OFO tools is required.The process follows this workflow:

  • Use Geograypher to (1) crop out each view of each tree from each drone photo it appears in and (2) link each crop to the ID of the corresponding detected tree.
  • Use the Tree Classification Framework to generate a prediction of species and live/dead status for each individual tree crop
  • Use the Tree Classification Framework to aggregate the predictions for each individual tree crop into a single prediction for each detected tree.

TODO: Consider moving the above text into the actual workflow sections below.

TODO: Consider adding illustrative schematics here or (more likely) below.

Required input data

Name Source Example file
Detected tree crowns Tree detection composite1_tree-crowns_geometric_v01.gpkg
Raw drone photos Drone flight raw-drone-photos/composite1
Photogrammetry camera locations Photogrammetry composite1_cameras.xml

The detected tree crowns may be ones you produced in the Detection task, or you can use the pre-generated detections in the tutorial data folder (linked above). It is important that the detections and photogrammetry camera locations be derived from the same photogrammetry run, using the same raw drone photos that you will use for this task.

Setup

Before you begin

Make sure you have gone through the OFO tutorial preparation guidance.

Follow our Docker guide to understand what Docker is and ensure your system is set up to run Docker containers.

TODO: Need to merge the install instructions for both tools here?

Install the Geograypher and Tree Classification Framework libraries and their dependencies into a Conda environment following their installation instructions, then activate the Conda environment in a terminal window:

conda activate TODO: Name of the Conda environment

TODO: Repeat the Conda env instructions from above once complete.

Then load the required libraries:

# TODO: Complete python lines to load libraries

Crop the view of each tree from each drone photo

Before we can use computer vision to classify trees to species, we need to:

  • Take the geospatial tree crowns that we detected in the Detection task
  • For detected tree crown, find all the drone photos it appeared in
  • For each photo the tree appeared in crop the view of it from the photo.

This means that ultimately, for every geospatial tree crown, we will end up with multiple cropped images, each from a different viewpoint.

TODO: Illustrative schematic here?