You are here

Home › LifeCLEF 2026

AnimalCLEF2026

AnimalCLEF

Discovery and Re-Identification of Individual Animals

Tentative Schedule

  • 17 November 2025: Registration opens for all LifeCLEF challenges (registration is free of charge)
  • 1 February 2026: Competition Start
  • 23 April 2026: Registration closes for all LifeCLEF challenges
  • 7 May 2026: Competition Deadline
  • 28 May 2026: Deadline for submission of working note papers by participants [CEUR-WS proceedings]
  • 30 June 2026: Notification of acceptance of working note papers [CEUR-WS proceedings]
  • 6 July 2026: Camera-ready deadline for working note papers.
  • 21-24 Sept 2026: CLEF 2026 Jena - Germany

All deadlines are at 11:59 PM CET on a corresponding day unless otherwise noted. The competition organizers reserve the right to update the contest timeline if they deem it necessary.

Motivation

Animal re-identification supports core wildlife objectives, e.g., estimating population sizes, monitoring movements, and analysing behaviour, by linking images to unique individuals. Automation scales these efforts by reducing manual matching and lowering annotation costs across species and sites, which is especially important when data volumes grow faster than expert capacity. Most existing benchmarks emphasise verification against a known catalogue (closed-set or open-set matching), which answers whether an image corresponds to a known individual but does not assign identities to previously unseen animals. In practice, building or updating a catalogue requires discovery: grouping images of unknown individuals so that each cluster corresponds to a single animal and can be incorporated into long-term monitoring workflows. This setting is challenging due to appearance variation (pose, illumination, ageing, injuries), background clutter, domain shifts across sensors and habitats, and class imbalance. By targeting the discovery step directly through unsupervised clustering of test images under realistic, multi-species, and variable conditions, the competition aims to advance methods that are immediately useful for creating and maintaining field-ready identity databases with minimal intervention.

Task Description

This challenge is all about the discovery of the following two species: (i) loggerhead sea turtles (Zakynthos, Greece), and (ii) Texas horned lizards (Texas, USA). Your goal will be to create clusters of images. Each cluster should correspond to one individual animal.

Participants may decide to work with the relatively small provided dataset or to boost the model performance by employing the WildlifeReID-10k dataset: a collection of 36 existing wildlife re-identification datasets, with additional processing and diverse species such as marine turtles, primates, birds, African herbivores, marine mammals, and domestic animals. WildlifeReID-10k contains approximately 140,000 images of over 10,000 individuals.

Data Collection

The objective of this competition is to develop a model capable of discovering individual animals from images. For feature extraction, we provide a pre-trained model MegaDescriptor, and for pre-training, we provide a large-scale dataset WildlifeReID-10k with 10k identities and around 140k images. Participants may use any re-identification dataset from the WildlifeDatasets package.

For testing, we have curated a dataset consisting of data of two species: (i) loggerhead sea turtles (Zakynthos, Greece), and (ii) Texas horned lizards (Texas, USA). The identities in the datasets were determined by field experts and validated by AI models.

More info on a Kaggle competition platform.

Evaluation process

Performance will be evaluated using the Adjusted Rand Index (ARI), which measures pairwise consistency between the predicted and true clusters. This metric penalizes both over-clustering (splitting one individual into multiple) and under-clustering (merging different individuals into a single one). The provided dataset is small, with a size of less than 1 GB. Optionally, participants may choose to use the WildlifeReID-10k dataset, which is approximately 25 GB in size, to enhance their training or evaluation protocol.

Rules

Using additional data or metadata is permitted on condition that for each run with external data, you submit an equivalent run with only the data supplied, to enable more accurate comparisons.

Self-supervised, semi-supervised or unsupervised approaches are strongly encouraged. Pre-trained models are provided.

Participation requirements

Publication Track

All registered participants are encouraged to submit a working-note paper to peer-reviewed LifeCLEF proceedings (CEUR-WS) after the competition ends.
This paper must provide sufficient information to reproduce the final submitted runs.

Only participants who submitted a working-note paper will be part of the officially published ranking used for scientific communication.

The results of the campaign appear in the working notes proceedings published by CEUR Workshop Proceedings (CEUR-WS.org).
Selected contributions among the participants will be invited for publication in the Springer Lecture Notes in Computer Science (LNCS) the following year.

Organizers

Credits

University of West Bohemia      ZCU-FAV      Queen Mary University of London     Inria

ImageCLEF - Cross-language image retrieval evaluations