ter Avest, E. | bioRxiv Science

bioRxiv2025

BirdAVES in the wild: individual recognition as a step toward zebra finch communication networks

Understanding who communicates with whom, when, and how is central to the ecology of group-living animals, yet individual-level acoustic identification of animals in their natural environment remains challenging. Zebra finches are a model species whose vocal behaviour has been predominantly studied indoors; here we address the outdoor setting and investigate bioacoustic deep learning for individual identification at scale as a key step to build communication networks from field recordings. We fine-tune BirdAVES for recognizing 173 zebra finch individuals from short (1-3 s) clips using a concise training recipe: two-phase training, weighted sampling and class-weighted cross-entropy for long-tailed counts, and a supervised contrastive term to pull same-individual embeddings together. On a real-world dataset (2,915 clips, 173 individuals), the selected model achieved macro-F1 = 0.733 (val) / 0.726 (test) and steep retrieval gains (Top-5 = 0.868, Top-10 = 0.893 on test set). This enables conversion of hours of audio into "who-sang-when" timelines. We deliberately report top-k performance because it quantifies review effort and supports human-in-the-loop workflows by shrinking the number of clips an expert must audit. While a train-val/test gap reflects short windows and class imbalance, the embeddings are discriminative and immediately useful. Key next steps are to address the imbalance in our data and scaling towards a significantly larger set of individuals, and to translate individual recognition into communication or social networks.

ecology↗

bioRxiv2025

Communication networks of wild zebra finches (Taeniopygia castanotis)

Communication networks are widespread across species, permitting information flow and facilitating social connections across space and time. In birds, communication networks are well studied in territorial species with long-range songs connecting individuals across space, where unintended listeners extract information from others signalling interactions. Yet, acoustic signals also play important social roles at short range, forming communication networks that connect individuals within larger social units. Wild zebra finches (Taeniopygia castanotis) provide a unique model system to examine such communication networks in a non-territorial species. Zebra finches breed in loose colonies in the Australian arid zone, move around in pairs or small groups, and gather at social hotspots, thus forming dynamic, potentially multi-level, societies. Here, we quantified singing activity and connectivity using the individually distinctive male song recorded at two breeding sub-colonies and three social hotspots over one to three days. We identified 1,835 song bouts from 163 males based on spectrographic similarities and we assessed within and between individual song assignments with a deep learning model (BirdNET). We constructed communication networks based on temporal singing proximity at shared locations: social hotspots and breeding colonies. We reveal higher singing activity at social hotspots than at breeding sites, with almost no dawn song at either site. Singing peaked later, yet at different times of day between breeding sites and social hotspots. Communication networks, with distinct males singing in close temporal proximity, were apparent in both contexts, with larger networks at hotspots. These networks included some individuals that sang together repeatedly at either site, but overall networks were not strongly nested, with only very few males maintaining associations across breeding colonies and hotspots. These networks may facilitate synchronised foraging and breeding as adaptations to a harsh and unpredictable environment. Additionally, our approach offers a novel road map for widening the understanding of communication networks.

animal behavior and cognition↗