Nearest-Class Mean and Logits Agreement for Wildlife Open-Set Recognition

2025-10-20

AI Summary

To address the overconfidence of closed-set models on unknown classes in wildlife classification, this paper proposes a post-hoc open-set recognition (OSR) method that requires no model retraining. The core innovation lies in establishing a consistency metric between feature space—built upon nearest-class-mean (NCM) embeddings—and logit space—characterized by softmax probabilities and raw logit distributions—where deviations between these spaces serve as discriminative signals for known versus unknown samples. Unlike existing OSR approaches that rely on task-specific fine-tuning or architectural modifications, our method operates entirely at inference time without altering model parameters. Evaluated on African and Swedish wildlife datasets, it achieves AUROC scores of 93.41% and 95.35%, respectively, demonstrating strong cross-dataset generalization and state-of-the-art overall performance—significantly outperforming mainstream methods optimized for single-dataset settings.

Technology Category

• Machine Learning: Multi-class/Multi-label Learning & Extreme Classification
• Computer Vision: Object Detection & Categorization
• Data Mining & Knowledge Management: Anomaly/Outlier Detection

Application Category

• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

Current state-of-the-art Wildlife classification models are trained under the closed world setting. When exposed to unknown classes, they remain overconfident in their predictions. Open-set Recognition (OSR) aims to classify known classes while rejecting unknown samples. Several OSR methods have been proposed to model the closed-set distribution by observing the feature, logit, or softmax probability space. A significant drawback of many existing approaches is the requirement to retrain the pre-trained classification model with the OSR-specific strategy. This study contributes a post-processing OSR method that measures the agreement between the models' features and predicted logits. We propose a probability distribution based on an input's distance to its Nearest Class Mean (NCM). The NCM-based distribution is then compared with the softmax probabilities from the logit space to measure agreement between the NCM and the classification head. Our proposed strategy ranks within the top three on two evaluated datasets, showing consistent performance across the two datasets. In contrast, current state-of-the-art methods excel on a single dataset. We achieve an AUROC of 93.41 and 95.35 for African and Swedish animals. The code can be found https://github.com/Applied-Representation-Learning-Lab/OSR.

Problem

• Classify known species while rejecting unknown wildlife samples
• Address overconfidence in predictions when encountering unknown classes
• Propose post-processing method without retraining existing classification models

Innovation

• Uses nearest-class mean distance for probability distribution
• Compares NCM distribution with softmax probabilities
• Post-processing method requiring no model retraining