> ## Content Index
> Fetch the complete content index at: https://intelligence.wildlandseco.com/llms.txt
> Use this file to discover other available public pages before exploring further.

# Small-mammal footprint AI is promising, not yet field-proven
- URL: https://intelligence.wildlandseco.com/2026-10-28-small-mammal-footprint-ai-validation/
- Published: 2026-10-28T13:33:56.000Z
- Updated: 2026-10-28T13:33:56.000Z
- Description: A trackplate classifier reported 89% grouped-validation accuracy on southern African mammals.
- Author: Benjamin Worlund
- Tags: GIS, Wildlife, Methods, #ecological-intelligence, #brief-2026-10-28, #weak-signal, #geo-africa

Evidence

Peer-reviewed early-release methods

Geography

Five sites, southern Africa

System

Small-mammal footprints and automated monitoring

## WHAT THE EVIDENCE SAYS

A vision model trained on 3,355 images from 122 trackplates for 11 species averaged 89.2% accuracy with five-fold cross-validation grouped by trackplate.

## WHY IT MATTERS / OUR READ

The next test is performance across new substrates, weather, sites and species, not a larger internal accuracy score.

## CAVEATS / UNCERTAINTY

Grouped internal cross-validation is not an independent field deployment or validated rare-species absence survey. Accepted Oct 27; final formatting pending.

## SOURCE

Frontiers in Ecology and Evolution — Liske et al. (accepted 27 Oct 2026)  
[View original evidence ↗](https://doi.org/10.3389/fevo.2026.1983895?ref=intelligence.wildlandseco.com)

From the Morning Brief · 2026-10-28

[Read the full Morning Brief →](https://intelligence.wildlandseco.com/morning-brief-2026-10-28/)