Accuracy and limitations

Honest expectations build better botanists—and safer landscapes

Probabilities, not oracles

Computer vision ranks candidates; it does not “know” a plant the way a taxonomist who has keyed it in the hand does. Treat every output as a hypothesis to weigh against field characters and references.

When mistakes hurt

Errors matter most around toxicity, invasives, protected species, compliance, and teaching. In those settings, obtain independent confirmation and document your reasoning.

What raises confidence

Known weak spots

Seedlings, highly degraded material, cultivars far from wild-type, and regions with thin training signal can all widen error bands. Hybrid swarms and apomictic complexes remain hard even for specialists.

Our commitment

We refine models and data continuously, publish limitations plainly, and welcome corrections that improve safety and science. Tell us what broke in the field.

FAQ · Terms of use