AI scans 22 million old plant specimens to spot flowers
Grady EL, LaFrance R, Li D, Dinnage R, Denny EG
Phenology
The pressed plant specimens sitting in museum collections for over a century are becoming a searchable record of how flowering times have shifted as the climate has warmed, letting researchers (and curious gardeners) see exactly when a species used to bloom versus when it blooms now.
Museums hold millions of dried, pressed plants collected over hundreds of years, and each one is a snapshot of when that plant was flowering at that moment in history. Checking all of them by hand for flowers would take forever, so researchers trained a computer model to look at photos of these specimens and automatically flag which ones show blooms. They ran it on 22 million specimen images, found flowers on 11.1 million of them, and are feeding that information into a public database so scientists can study how bloom timing has changed over the decades.
Key Findings
The ensemble machine learning pipeline detected floral structures with relatively strong accuracy but had moderately high false negative rates, meaning it sometimes missed flowers that were actually present.
Out of 22 million filtered herbarium specimen images, 11.1 million were labeled as containing flowers.
Only 2.9 million of the flower-labeled records had complete metadata usable for phenology research, revealing a major bottleneck in label digitization.
chevron_right Technical Summary
Scientists built an AI system that scans millions of digitized herbarium plant specimens to automatically detect which ones have flowers, creating a massive new dataset for tracking how flowering times have shifted over history.
Abstract Preview
Original paper
Petal to the metal: The slow road to automating large-scale phenology labeling for herbarium specimens.
Herbarium specimens represent critical historical records of plant phenology, yet automating annotation of reproductive structures remains challenging given the diversity of floral morphologies, sp...
open_in_new Read full abstractAbstract copyright held by the original publisher.
Was this useful?
Want to tell us more? (optional)
Thanks for the note!
Something went wrong — please try again.
Too many submissions. Try again in an hour.
Temperature regulation in plants: From molecular mechanisms to climate-resili...
The wheat in your bread and the vegetables in your garden depend on precise temperature cues to know when to flower and set seed — and scientists decoding ex...
Climate adaptation in plants refers to the physiological and evolutionary mechanisms through which plants adjust to changing environmental conditions, including temperature shifts, altered precipitation patterns, and seasonal variations. Understanding these processes is essential for plant science
arrow_forward Explore topic