Tasks
We are excited to announce the 2026 BioDCASE task line up!
Multi-Channel Alignment
Researchers often deploy multiple audio recorders simultaneously, for example with passive automated recording units (ARU's) or embedded in animal-borne bio-loggers. Analysing sounds simultaneously captured by multiple recorders can provide insights into animal positions and numbers, as well as the dynamics of communication in groups. However, many of these devices are susceptible to desynchronization due to nonlinear clock drift, which can diminish researchers' ability to glean useful insights. Therefore, a reliable, post-processing-based re-synchronization method would increase usability of collected data.
In this challenge, participants will be presented with pairs of temporally desynchronized recordings and asked to design a system to synchronize them in time. In the development phase, participants will be provided audio pairs and a small set of ground-truth synchronization key points--the likes of which could be produced by a manual review of the data. In the evaluation phase, participants' systems will be ranked by their ability to synchronize unseen audio pairs.
Organisers
Supervised detection of strongly-labelled Antarctic blue and fin whale calls
Passive Acoustic Monitoring (PAM) is a technology used to analyze sounds in the Ocean, where our capacity of visual observation is highly limited. It has emerged as a transformative tool for applied ecology, conservation and biodiversity monitoring. In particular, it offers unique opportunities to examine long-term trends in population dynamics, abundance, distribution and behaviour of different whale species. But for this purpose, the automation of PAM data processing, involving the automatic detection of whale calls in long-term recordings, faces two major issues: the scarcity of calls and the variability of soundscapes.
In this data challenge, a supervised sound event detection task was designed, and applied to the detection of 7 different call types from two emblematic whale species, the Antarctic blue and fin whales. This task aims to improve and assess the ability of models to address the two issues just mentioned, as models will have to deal with whale calls happening only 6 % of the time, and PAM recordings coming from different time periods and sites all around Antarctica that present highly variable soundscapes. The White Continent appeared to be a very exciting playground to start a large-scale evaluation of model generalization capacity, but challenging for sure!
Organisers
Paul Carvaillo
France Energies Marines
Anatole Gros-Martial
Centre d’Etudes Biologiques de Chizé, GEO-Ocean
Lucie Jean-Labadye
Sorbonne Université, LAM
Brian Miller
Australian Antarctic Division
Paul Nguyen Hong Duc
Curtin University
Pierre-Yves le Rolland Raumer
IUEM
Bioacoustics for Tiny Hardware
The next generation of autonomous recording units contains programmable chips, thus offering the opportunity to perform BioDCASE tasks. On-device processing has multiple advantages, such as high durability, low latency, and privacy preservation. However, such “tiny hardware” is limited in terms of memory and compute, which calls for the development of original methods in audio content analysis. In this context, task participants will revisit the well-known problem of automatic detection of birdsong while adjusting their systems so as to meet the specifications of a commercially available microcontroller.
Organisers
Christian Walter
Active Learning for Bioacoustics
Active
A fundamental challenge across bioacoustics domains (terrestrial and marine) is the annotation of unlabelled data. Passive acoustic monitoring systems generate vast amounts of data, but only a small portion can be feasibly annotated by expert human annotators. Since model performance depends heavily on the quality and quantity of labelled data, this raises the following research question: Given vast amounts of raw acoustic data and limited annotation resources, which data should be prioritised for labelling?
Active learning (AL) is a critical strategy for scaling bioacoustic monitoring. AL is an iterative method of data selection, annotation and model training also often within a human-in-the-loop framework. Fundamentally, AL aims to optimise for a learning objective (e.g. model performance) using less labeled data minimising annotation requirements. Participants will design an active learning strategy (acquisition function) to maximise training efficiency across batches of multi-label data considering informativeness quantification, diversification, long-tail performance and cross-domain generalisation.
Organisers
Cross-Domain Mosquito Species Classification
Mosquito Classification
Mosquito-borne diseases affect over a billion people each year and cause close to one million deaths. Traditional mosquito surveillance relies on traps and manual identification; this process is slow, labour-intensive, difficult to scale, and can expose researchers to infection risk. Mosquito species classification (MSC) from bioacoustic data offers a practical alternative, using low-cost sensors (such as phones) to identify species from flight sounds, enabling more widespread monitoring where traditional mosquito surveillance is more difficult to implement.
In practice, mosquito flight tones are often faint and easily masked by background noise. Furthermore, mosquito species composition differs across regions, and intrinsic variability within species (size, sex, age) combined with external variables (e.g., temperature, different recording devices) means even the same species' flight tones span a broad range of fundamental frequencies. These factors lie at the heart of this challenge. Our goals are to learn from such challenging data, achieve robustness to real-world variability, and generalise across domains. In this challenge, a domain refers to a specific set of acquisition conditions (e.g., location, device, environment). Participants are invited to train their models on recordings from multiple source domains and evaluate performance on unseen domains that differ by geography, device type, or acoustic environment, reflecting realistic deployment conditions.
Organisers
Bird Counting
Counts
Estimating the number of individual birds from acoustic recordings is a fundamental challenge in biodiversity monitoring. This task addresses bird abundance estimation in zoo aviaries with known ground-truth population counts. Participants will receive collections of ~20,000 short audio fragments extracted from continuous recordings in multi-species aviaries containing known numbers of a target species, alongside other bird species. The recordings capture birds vocalizing naturally in groups over extended periods, creating realistic acoustic complexity including overlapping vocalizations, environmental noise, and natural behavioral variation.
The task is to estimate the number of individuals of the target species in each collection. The main leaderboard ranks systems based solely on target-species abundance estimation performance. Participants may optionally extend their methods to additional species; a secondary leaderboard showcases generalist systems estimating multiple species, but does not contribute to final rankings. Species detection itself is not evaluated; systems may use any detection strategy to focus on the target species within the multi-species environment. Performance is assessed solely on the final group-size estimates. Success on this task has direct applications to passive acoustic monitoring where traditional surveys are impractical.
