WildID › Place lists and GPS

Place lists and GPS

A place list limits the candidates to the species recorded in the country or region where you took the photo. It is one of the two settings that improve the answers most. The other is the crop.

Why a list helps

Without a list the model chooses among 867,455 names from the whole world. With the list for Kenya it chooses among about 1,400, and with the list for Madagascar among about 860. Look-alike species from other continents are no longer candidates.

On 401 of our lemur photos the Madagascar list raised the share of correct names from 69.8 % to 77.1 % with BioCLIP 2. BioCLIP 2.5 gained less, from 79.8 % to 80.8 %.

Where the lists come from

A list contains the species that observers on iNaturalist have recorded in the place, counting research-grade observations only. By default it covers mammals, reptiles, amphibians and birds. Birds are included on purpose: without them a bird photo would be forced into a mammal or a reptile.

iNaturalist and the model sometimes spell a species differently. WildID looks such names up once in the GBIF Backbone Taxonomy and remembers the result. A list is fetched once, stored on your computer and used offline from then on.

Fetching a list in Lightroom

Choose Library → Plug-in Extras → WildID: prefetch place list… and fill in the place:

Country
The country name in English, for example Kenya.
Region
Optional. A state or province inside the country, for example Mato Grosso in Brazil.
Place id
Optional, for any other iNaturalist place such as a national park. The id is the number after place_id= in the address bar when you filter observations by that place on inaturalist.org.
Taxa
The groups to include. The default is Mammalia,Reptilia,Amphibia,Aves.
Refresh even if already cached
Fetches the list again, for example before a new trip to the same country.

Fetching is the only step that uses the internet after the setup. A large country can take a few minutes.

Using a list when you identify

The Candidates menu in the Identify dialog has three settings:

  • The full TreeOfLife: no list. Every name the model knows is a candidate.
  • A prefetched place: type the country, the region or the place id exactly as you fetched it.
  • Each photo’s GPS position: see below.

Next to the taxa you can choose between “species list” and “whole genera”. Whole genera keeps every species of each genus on the list. It is meant for places with few iNaturalist observations and it makes the candidate list several times longer.

A list for every photo, from GPS

If your photos carry a position, WildID can pick the country or the region for each photo by itself. It uses the position stored in the Lightroom catalog, so positions you set in the Map module count as well as those from the camera.

  1. Select the photos and open WildID: prefetch place list…
  2. Under For, choose the selected photos’ GPS positions, pick the GPS level, country or region, and click Fetch. WildID looks up one position for each area of about 28 by 28 km and fetches the lists for the places it finds.
  3. In the Identify dialog, set Candidates to each photo’s GPS position and choose the same GPS level.

A photo without a position is identified without a list and is not flagged. A photo in an area you did not prefetch is also identified without a list, and it is flagged for review. The summary counts both.

The safety net

A list can hide the right answer. For that reason WildID also asks the model for its answer without the list. If that answer is not on the list, the photo is flagged for review.

Limits

  • Animals in zoos and parks, and escaped or introduced species, are usually missing from the list of the country they were photographed in.
  • Rare visitors and species that few people record on iNaturalist can be missing as well.
  • A list is as good as the observations behind it. For well-visited countries it is thorough, for remote regions it can be thin. Whole genera helps there.

On the command line

uv run wildid prefetch --country Kenya
uv run wildid prefetch --country Brazil --region "Mato Grosso"
uv run wildid photos/ --country Kenya
uv run wildid prefetch --gps-from photos/
uv run wildid photos/ --geo-auto

Previous: When a name is wrongNext: What it writes to Lightroom