Unlocking AI as a research tool for plant microscopy
Software created by University of Queensland scientists could help AI platforms overcome a blind spot limiting their value in plant research. UQ Ph.D. student Tianqi Wei said even advanced vision-language models (VLMs) had difficulty interpreting images of microscopic plant details.
University of Queensland researchers have developed a new AI benchmarking dataset called PlantMicro that could help overcome limitations in using AI platforms for plant microscopy research. PhD student Tianqi Wei explained that even advanced vision-language models (VLMs) have difficulty interpreting microscopic plant details, which hinders their value in plant research.
PlantMicro, now available on GitHub, integrates thousands of microscopic images and uses sophisticated training mechanisms to test VLMs' ability to identify, sort and count minute plant features and diseases. The software was presented at the European Conference on Computer Vision (ECCV 2026) in Sweden and was partly supported by the Grains Research and Development Corporation (GRDC).
Experiments with PlantMicro showed that both closed- and open-source VLMs performed poorly, achieving average accuracies of about 30%, which is only slightly better than random guessing. This indicates that VLMs still face significant challenges in accurately identifying plant hosts and diseases from microscopic evidence. The researchers hope that PlantMicro will provide a better foundation for AI to be used in plant science, ultimately contributing to food security and environmental sustainability.
Written by urgent.news from Phys.org's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
