Benchmarking NACTI Species Recognition in Long-Tailed Regimes

2026-07-20

AI Summary

This study addresses the performance imbalance in species recognition caused by the long-tailed distribution in the North American Camera Trap Image (NACTI) dataset. Building upon the PyTorch Wildlife foundation model, the authors systematically evaluate various long-tailed recognition (LTR) approaches, including specialized loss functions and LTR-aware regularization strategies. The work establishes the first LTR benchmark on a large-scale, real-world camera trap dataset and introduces three low-bias test sets—ENA-Detection, Caltech, and Missouri—to assess out-of-domain generalization. Experimental results demonstrate that the optimized model achieves a Top-1 accuracy of 99.40% on NACTI, substantially outperforming the baseline, and exhibits enhanced robustness on out-of-domain evaluations. Nevertheless, the study also reveals that existing LTR methods still suffer from prediction collapse for rare species under extreme domain shifts.

Abstract

As with most ``in the wild'' collections of the natural world, the North America Camera Trap Images (NACTI) dataset exhibits long-tailed class imbalance, with the largest class covering over 50% of its 3.7M images. Building on the PyTorch Wildlife model, we systematically evaluate Long-Tail Recognition (LTR) methodologies to benchmark species recognition performance, including specialised loss functions and LTR-sensitive regularisation. Our optimised configuration achieves state-of-the-art 99.40% Top-1 accuracy on the NACTI test split, significantly outperforming standard baselines and previously reported top performances. To assess robustness under domain shifts (e.g., night-time captures, occlusion, motion-blur), we extend our evaluation across three independent reduced-bias test sets (including ENA-Detection, Caltech Camera Traps and Missouri Camera Traps). Across these out-of-distribution (OOD) evaluations, our LTR-enhanced model consistently demonstrates substantially stronger generalisation capabilities compared to standard cross-entropy approaches. However, qualitative and quantitative analyses underline that current LTR optimisations cannot fully overcome representational bottlenecks, resulting in catastrophic predictive breakdown for rare `Tail' classes under severe domain shift. For maximum reproducibility, all dataset splits, key code, and network weights are published with this paper at https://github.com/ZehuaLiuY/Species-Classification.

Problem

• long-tailed recognition
• class imbalance
• species recognition
• domain shift
• camera trap images

Innovation

• Long-Tail Recognition
• Species Recognition
• Domain Shift Robustness
• Camera Trap Images
• Class Imbalance