Researchers from the Mediterranean Institute for Advanced Studies (IMEDEA), a joint institute of the CSIC and the University of the Balearic Islands, have developed an artificial intelligence system capable of identifying individual Balearic wall lizards from photographs.

The system, called ReMatch, analyses distinctive patterns in an animal's scales and compares them with photographs already stored in a reference database. It can determine whether a photographed lizard is likely to be an individual already recorded or potentially a new animal.

The technology could give researchers a new way to monitor wild populations over many years without having to repeatedly capture and physically mark individual animals.

A Natural Pattern That Works Like a Fingerprint

Every Balearic wall lizard, Podarcis lilfordi, has a distinctive arrangement of scales on the underside of its body.

ReMatch uses this natural pattern as a biometric feature. Rather than relying on an artificial tag or physical marking, the system analyses points and lines within the scale pattern and compares them between photographs.

The researchers describe this as an opportunity to recognise the same individual across different observations, potentially allowing animals to be followed over long periods.

The “DNI” comparison therefore refers to the distinctive visual pattern used to distinguish individuals, rather than an actual identification document.

Up to 98.4% Correct Identification

Testing with Balearic wall lizards produced particularly strong results.

When the photographed animal was already represented in the reference database, ReMatch identified the correct individual as its first choice in 98.4% of cases.

A second test included animals that had not previously been registered. In that open-set scenario, the system achieved an accuracy rate of 92.46%.

The distinction between these two tests is important. The first measures how well the system can identify an animal already known to the database, while the second also tests whether it can recognise that a photographed animal may not belong to the existing group of identified individuals.

Thousands of Photographs From Mallorca

The Balearic wall lizard was one of the main datasets used to develop and evaluate the system.

Researchers worked with 4,619 photographs representing 1,009 individual lizards. The images were collected during long-term fieldwork on Illot d'en Curt, a small islet off the southern coast of Mallorca, between October 2010 and September 2024.

The resulting BalearicLizard dataset has also been made publicly available for research and computer-vision applications.

The scientific dataset paper describes it as a curated collection designed to support non-invasive individual re-identification, ecological monitoring and computer-vision research.

Following Lizards Without Capturing Them

Being able to recognise an individual animal from photographs could significantly simplify long-term wildlife monitoring.

Traditional monitoring can involve capturing animals and applying tags, rings or other identifying marks before releasing them. A photographic identification system provides another option when the species has sufficiently distinctive natural patterns.

Researchers can potentially compare a photograph taken today with images collected years earlier to establish whether the same animal has been observed previously.

That information can contribute to studies of survival, reproduction, population size and changes within wild populations.

The approach is particularly useful for species that are difficult to monitor continuously in the field.

ReMatch Can Work With Other Species

The research team did not develop ReMatch exclusively for lizards.

The system was tested using several different types of natural patterns, including the stripes of plains zebras, the fur patterns of Saimaa ringed seals and human fingerprints. It was also tested on whale sharks, where its performance was weaker.

This suggests that the underlying approach could have applications beyond reptiles, particularly for species with detailed, relatively stable local patterns.

The researchers say the system performs especially well when the identifying features are fine-scale and local, while major changes in viewing angle or photographic conditions can make identification more difficult.

An Open-Set Approach

One of ReMatch's useful features is that it does not assume every photographed animal must already exist in the database.

If none of the stored identities reaches a sufficiently high confidence level, the system can flag the photograph as potentially belonging to a previously unrecorded individual.

That is particularly important in wildlife research, where scientists regularly encounter animals that have never been photographed before. The new individual can subsequently be added to the reference collection.

From Research Data to Conservation

The technology forms part of a wider effort to make use of the growing amount of photographic information collected during ecological fieldwork.

For researchers, being able to identify individual animals without repeatedly handling them could make long-term monitoring more practical and less invasive.

The Balearic wall lizard dataset also provides researchers with a substantial photographic resource for developing and testing computer-vision techniques.

For Ibiza and the wider Balearic Islands, the research is particularly relevant because the Balearic wall lizard is an emblematic island species whose populations can be studied using increasingly sophisticated non-invasive methods.

A New Tool for Understanding Wildlife

ReMatch does not replace fieldwork or ecological expertise. Instead, it provides researchers with another tool for determining whether photographs taken at different times may show the same individual.

Its results with Balearic wall lizards demonstrate the potential of combining long-term ecological datasets with artificial intelligence.

With thousands of photographs already available and the technology capable of recognising both known animals and potential newcomers, researchers now have another way to study individual lizards and the populations they form over time.

ReMatch and Balearic Lizards: Key Facts

• System: ReMatch

• Researchers: IMEDEA, CSIC and University of the Balearic Islands

• Species tested: Balearic wall lizard, Podarcis lilfordi

• Identification feature: Natural pattern of ventral scales

• Known-individual identification: 98.4% correct as first choice

• Open-set test: 92.46% accuracy

• Dataset: 4,619 photographs

• Individual lizards represented: 1,009

• Fieldwork location: Illot d'en Curt, Mallorca

• Collection period: October 2010 to September 2024

• Dataset name: BalearicLizard

• Main potential application: Non-invasive individual and population monitoring