THE AI STILL NEEDS YOUR TREE PHOTOS.
AI-generated images helped species-identification models when real photos were scarce — but actual photographs from people in the real world still performed better. Your random tree picture remains scientifically employed.
THE AI STILL NEEDS YOUR TREE PHOTOS
A new NC State study tested whether AI-generated images can stand in for real photographs when training species-identification models. Short answer: they can help at the margins, and they cannot replace you.
WHAT HAPPENED
Researchers at NC State's Center for Geospatial Analytics, led by Thomas Lake with co-author Chris Jones, generated synthetic images of 20 common North American trees and mixed them into training data alongside real photographs from iNaturalist and the Auto Arborist Dataset.
Synthetic imagery improved model performance specifically where real images were scarce. Where real photographs were plentiful, they outperformed the synthetic ones, and the authors are explicit that AI-generated imagery supplements rather than replaces field observations and community science.
The paper appears in Remote Sensing in Ecology and Conservation.
WHY THIS IS HOPESLOP
Every other week something insists that the machines no longer need us. Here is a study politely establishing the opposite, in the least dramatic way possible.
The synthetic pictures are a patch for gaps in the record. The record itself is still made of people standing in front of trees, taking a slightly blurry photo. That job is safe.
ORIGINAL SOURCENC State UniversityRead the original report