Related Experiment Video
Updated: Jun 24, 2026
08:47
Computer Vision-Based Biomass Estimation for Invasive Plants
Published on: February 9, 2024
Investigating key drivers influencing AI-based detection and identification of plants
Andréanne Charron1, Adèle Julien2, Joseph R Stinziano3
1Genotyping-Botany Laboratory, Canadian Food Inspection Agency, Ottawa, Ontario, Canada.
Plos One
|March 2, 2026
Summary
Restricting location data significantly impairs iNaturalist's ability to detect invasive alien plants (IAP). This highlights potential missed detections, emphasizing the need to consider location for effective biological monitoring and early IAP detection.
Area of Science:
- Ecology
- Botany
- Computational Biology
- Citizen Science
Background:
- AI-driven citizen science platforms are crucial for biological data collection, often using visual and geospatial data for species identification.
- Invasive alien plants (IAP) pose significant ecological and economic threats, necessitating accurate and timely detection methods.
- The accuracy of AI-based identification tools like iNaturalist can be influenced by various factors, including location data.
Purpose of the Study:
- To investigate the impact of location parameters on iNaturalist's plant identification accuracy, specifically for invasive alien plants (IAP).
- To compare the identification accuracy of iNaturalist and PlantNet, exploring potential biases in their performance.
- To assess how factors like plant family and visible plant parts affect identification accuracy.
Main Methods:
- Photographs of established plants (native/naturalized in Ontario, n=61) and "outsider plants" (regulated pests with limited distribution in Canada, n=402) were analyzed.
- iNaturalist and GBIF datasets were utilized for "outsider plant" photographs.
- A scoring system and cumulative linked mixed model were applied to analyze taxonomic accuracy, considering plant families, distribution status, and visible parts.
Main Results:
- Restricting location data significantly reduced iNaturalist's accuracy in identifying invasive alien plants (IAP), increasing the risk of missed detections.
- Identification accuracy was significantly lower for species in the Poaceae family and for photographs displaying only leaves.
- While sample size limited a robust comparison, preliminary findings suggest differences in accuracy between iNaturalist and PlantNet.
Conclusions:
- Location data is a critical factor influencing the effectiveness of AI tools like iNaturalist for invasive alien plant (IAP) monitoring and early detection.
- The findings underscore the importance of optimizing AI identification algorithms to account for location-specific data to improve IAP surveillance.
- Further research is needed to comprehensively compare the accuracy of different AI identification platforms under various conditions.
