PubMed · 2025-07-03
This review shows how museums' pressed-plant collections, now digitized and freely available online, are becoming a powerful training resource for AI that can identify plants and answer botanical questions. The standardized format and expert-verified labels of herbarium specimens give AI models a quality advantage that field photos rarely match.
Digitized herbarium specimens offer a key advantage for AI training: expert-verified identifications tied to standardized taxonomies, yielding high label accuracy.
The physical standardization of herbarium sheets (consistent scale, flattened form, color-corrected imaging) reduces image variability that typically hampers computer-vision model performance.
The authors flag a decline in taxonomic and specimen-based botanical expertise as a driving reason to develop AI tools that can augment human capacity in plant identification and research.