Skip to content

Deep Learning-based QAQC tools for FAO's Forest Resource Assessment (FRA) Remote Sensing Survey (RSS)

License

Notifications You must be signed in to change notification settings

sig-gis/fao-models

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

85 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

fao-models

image image

image

Image Classification models for the FAO's Forest Resource Assessment data collection campaigns in Collect Earth Online.

The FAO FRA team is conducting its next round of global data collection surveys. They use Collect Earth Online and local trained interpreters. Quality of interpreter answers have been a concern in the past. Answers between multiple interpreters on the same plot may disagree for a variety of reasons, but there has not been a good way to flag certain plots or interpreters in an informed way. We are developing EO models to act as a 'source of truth' in a new CEO QA/QC feature, so that project admins can identify plots or interpreters who display high levels of disagreement with a 'source of truth' and/or other interpreters.

This particular repo (see its cousin sig-ssl4eo) trains custom deep learning models for our specific task: image classification. That is, a single label for a given image. Our current fine-tuned model is just for forest/non-forest classification, but there is potential for further development beyond a binary forest/non-forest classification.

Setup

Inference

Creating Configs

Training a Model

About

Deep Learning-based QAQC tools for FAO's Forest Resource Assessment (FRA) Remote Sensing Survey (RSS)

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published