Spore it Out: A Deep Learning Framework for Identifying Fungi from Spore Morphology to Bridge Biodiversity Knowledge Gaps.

Fachada, Vasco; Turner, Fraser; Druzhinina, Irina

Spore it Out: A Deep Learning Framework for Identifying Fungi from Spore Morphology to Bridge Biodiversity Knowledge Gaps.

Resource type

ConferenceItem Conference poster (published)

Creator (person)

Fachada, Vasco

ORCID

Turner, Fraser

Druzhinina, Irina

ORCID

Date published

March 10, 2026

Abstract

Fungi are fundamental to ecosystem health, yet remain a monumental blind spot in biodiversity conservation, severely underrepresented in monitoring and policy frameworks like the Kunming-Montreal Global Biodiversity Framework. A primary bottleneck is the lack of rapid, reliable, and scalable identification tools. DNA metabarcoding is costly and often fails with degraded material, historical collections or paleoecological samples, where a vast repository of fungal diversity data lies locked away. We propose a transformative AI-driven solution to this crisis. With taxonomy and computer science expertise, we are developing deep learning models for the automated identification of fungal species based on spore morphology. Moving beyond simple silhouette analysis, our computer vision approach is the first to integrate critical taxonomic characters like ornamentation, wall thickness, and chromatism. This aims to create a high-resolution "sporal fingerprint" for species identification. Using the hyper-diverse and spore-rich genus Russula as a proof-of-concept, we are building a model with data from Kew's unparalleled fungarium and new data from field mycology efforts. This interdisciplinary approach, co-developing tools with taxonomists, ensures the models are biologically meaningful and directly applicable to real-world conservation challenges. This tool will enable the identification of both extant and extinct species from modern collections to ancient sediments, where durable spores outlast degradable DNA. This innovation provides a scalable method to harmonize fungal monitoring across ecosystems and geographies, unlocking historical and forthcoming data for trend analysis. By generating rapid, cost-effective identifications, it aims to empower researchers, citizen scientists and industry players, fostering inclusive and collaborative dynamics. The resulting open-access database and toolset will serve as a foundational resource for upscaling fungal conservation, providing the actionable data needed to integrate fungi into national Red Lists, biodiversity action plans, and Key Biodiversity Areas (KBAs). Integrating this tool into established conservation programs adds the crucial fungal dimension, ensuring that fungi are no longer the overlooked majority, but a central part of the transformation towards biodiversity-positive futures.