Multispecies Animal Re-ID Using a Large Community-Curated Dataset

TL;DR

The paper presents MiewID, a multispecies animal re-identification model trained on a large, community-curated dataset spanning 49 species and 37k individuals. By using a single EfficientNetV2-based embedding trained with a sub-center ArcFace loss, the model consistently outperforms per-species baselines and demonstrates strong zero-shot and fine-tuning performance for new species. It also outperforms MegaDescriptor on unseen species and shows effective bootstrapping with limited data, supported by production deployment and open-source releases. The work highlights the practical benefits of cross-species training for wildlife monitoring and suggests that frequent retraining with expanding datasets can sustain improved identification performance at scale.

Abstract

Recent work has established the ecological importance of developing algorithms for identifying animals individually from images. Typically, a separate algorithm is trained for each species, a natural step but one that creates significant barriers to wide-spread use: (1) each effort is expensive, requiring data collection, data curation, and model training, deployment, and maintenance, (2) there is little training data for many species, and (3) commonalities in appearance across species are not exploited. We propose an alternative approach focused on training multi-species individual identification (re-id) models. We construct a dataset that includes 49 species, 37K individual animals, and 225K images, using this data to train a single embedding network for all species. Our model employs an EfficientNetV2 backbone and a sub-center ArcFace loss function with dynamic margins. We evaluate the performance of this multispecies model in several ways. Most notably, we demonstrate that it consistently outperforms models trained separately on each species, achieving an average gain of 12.5% in top-1 accuracy. Furthermore, the model demonstrates strong zero-shot performance and fine-tuning capabilities for new species with limited training data, enabling effective curation of new species through both incremental addition of data to the training set and fine-tuning without the original data. Additionally, our model surpasses the recent MegaDescriptor on unseen species, averaging an 19.2% top-1 improvement per species and showing gains across all 33 species tested. The fully-featured code repository is publicly available on GitHub, and the feature extractor model can be accessed on HuggingFace for seamless integration with wildlife re-identification pipelines. The model is already in production use for 60+ species in a large-scale wildlife monitoring system.

Figures (12)

• Figure 1: Comparison of top-1 performance of multi-species and single-species models. Species are ordered by increasing numbers of sightings.
• Figure 2: Top-5 performance comparison per species for model trained on the full dataset and model trained on all species except the test species
• Figure 3: Mean top-1 accuracy and standard deviation across species as a function of the maximum number of annotations per individual in the database.
• Figure 4: Comparison of Swin-V2 and Efficientnet-V2 backbones for MiewID training
• Figure 5: Comparison of top-1 performance of multi-species and single-species models. Species are ordered by increasing numbers of sightings.