Benchmarking NACTI Species Recognition in Long-Tailed Regimes
2026-07-20 • Computer Vision and Pattern Recognition
Computer Vision and Pattern Recognition
AI summaryⓘ
The authors studied a large set of animal photos taken in the wild, which has many more pictures of some animals than others. They tested different ways to improve recognizing rare species in these unbalanced datasets and made a new model that works very well, achieving high accuracy. They also checked how well the model handles tough cases, like night photos or blurry images, and found it does better than standard methods but still struggles with very rare species in difficult conditions. They shared their data and code to help others reproduce their results.
long-tailed class imbalancespecies recognitioncamera trap imageslong-tail recognitionout-of-distribution evaluationPyTorch Wildlife modelcross-entropy lossregularizationdomain shiftTop-1 accuracy
Authors
Zehua Liu, Tilo Burghardt
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.