Open-Set Recognition of Novel Species in Biodiversity Monitoring

2025-03-03

AI Summary

Open-set recognition—particularly the detection of previously unrecorded species—remains a critical challenge in biodiversity monitoring. Method: We introduce Open-Insects, the first fine-grained open-set benchmark for insects, comprising 59 field-collected putative novel species as out-of-distribution test samples. We propose a geography-aware open-set evaluation paradigm, enabling systematic assessment across geographically stratified difficulty levels. We further investigate lightweight post-hoc Softmax thresholding for novelty detection and analyze how auxiliary cues (e.g., geographic coordinates, morphological traits) enhance few-shot novel-species identification. Finally, we jointly optimize out-of-distribution detection, training regularization, and data augmentation. Contribution/Results: Our approach significantly improves novel-species detection rates, establishes a reproducible benchmark for AI-driven species discovery, and provides a practical technical framework for scalable biodiversity monitoring.

Technology Category

• Data Mining & Knowledge Management: Anomaly/Outlier Detection
• Search and Optimization: Sampling/Simulation-based Search
• Machine Learning: Multi-class/Multi-label Learning & Extreme Classification

Application Category

• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets
• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• Security and Privacy: Data transparency and provenance

Abstract

Machine learning is increasingly being applied to facilitate long-term, large-scale biodiversity monitoring. With most species on Earth still undiscovered or poorly documented, species-recognition models are expected to encounter new species during deployment. We introduce Open-Insects, a fine-grained image recognition benchmark dataset for open-set recognition and out-of-distribution detection in biodiversity monitoring. Open-Insects makes it possible to evaluate algorithms for new species detection on several geographical open-set splits with varying difficulty. Furthermore, we present a test set recently collected in the wild with 59 species that are likely new to science. We evaluate a variety of open-set recognition algorithms, including post-hoc methods, training-time regularization, and training with auxiliary data, finding that the simple post-hoc approach of utilizing softmax scores remains a strong baseline. We also demonstrate how to leverage auxiliary data to improve the detection performance when the training dataset is limited. Our results provide timely insights to guide the development of computer vision methods for biodiversity monitoring and species discovery.

Problem

• Develop methods for detecting new species in biodiversity monitoring.
• Evaluate algorithms for open-set recognition in varying geographical contexts.
• Improve detection performance using auxiliary data with limited training datasets.

Innovation

• Open-Insects dataset for open-set recognition
• Post-hoc methods using softmax scores baseline
• Auxiliary data improves limited training datasets