1. Ecological data are collected over vast geographic areas using digital sensors such as camera traps and bioacoustic recorders. Camera traps have become the standard method for surveying many terrestrial mammals and birds, but camera trap arrays often generate millions of images that are time-consuming to label. This causes significant latency between data collection and subsequent inference, which impedes conservation at a time of ecological crisis. Machine learning algorithms have been developed to improve the speed of labelling camera trap data, but it is uncertain how the outputs of these models can be used in ecological analyses without secondary validation by a human.
2. Here, we present our approach to developing, testing and applying a machine learning model to camera trap data for the purpose of achieving fully automated ecological analyses. As a case-study, we built a model to classify 26 Central African forest mammal and bird species (or groups). The model generalizes to new spatially and temporally independent data (n = 227 camera stations, n = 23,868 images), and outperforms humans in several respects (e.g. detecting ‘invisible’ animals). We demonstrate how ecologists can evaluate a machine learning model's precision and accuracy in an ecological context by comparing species richness, activity patterns (n = 4 species tested) and occupancy (n = 4 species tested) derived from machine learning labels with the same estimates derived from expert labels.
3. Results show that fully automated species labels can be equivalent to expert labels when calculating species richness, activity patterns (n = 4 species tested) and estimating occupancy (n = 3 of 4 species tested) in a large, completely out-of-sample test dataset. Simple thresholding using the Softmax values (i.e. excluding ‘uncertain’ labels) improved the model's performance when calculating activity patterns and estimating occupancy but did not improve estimates of species richness.
4. We conclude that, with adequate testing and evaluation in an ecological context, a machine learning model can generate labels for direct use in ecological analyses without the need for manual validation. We provide the user-community with a multi-platform, multi-language graphical user interface that can be used to run our model offline.

中文翻译：

---

机器学习模型标记的相机陷阱数据的稳健生态分析

1. 生态数据是使用数字传感器（例如相机陷阱和生物声学记录器）在广阔的地理区域收集的。相机陷阱已成为调查许多陆地哺乳动物和鸟类的标准方法，但相机陷阱阵列通常会生成数百万张图像，标记起来非常耗时。这会导致数据收集和后续推断之间存在显着延迟，从而在生态危机时期阻碍保护。已经开发出机器学习算法来提高标记相机陷阱数据的速度，但不确定如何在没有人工二次验证的情况下将这些模型的输出用于生态分析。
2. 在这里，我们展示了我们开发、测试和应用机器学习模型到相机陷阱数据的方法，以实现全自动生态分析。作为案例研究，我们建立了一个模型来对 26 种中非森林哺乳动物和鸟类（或群体）进行分类。该模型推广到新的空间和时间独立数据（n  = 227 个摄像站，n  = 23,868 张图像），并在几个方面优于人类（例如检测“隐形”动物）。我们展示了生态学家如何通过比较物种丰富度、活动模式（ 测试的n = 4 个物种）和占有率（n = 测试的 4 个物种）来自机器学习标签，与来自专家标签的估计相同。
3. 结果表明， 在大型完全样本外测试中计算物种丰富度、活动模式（ 测试的n = 4 个物种）和估计占用率（测试的 4 个物种中的n = 3 个）时，全自动物种标签可以等同于专家标签数据集。在计算活动模式和估计占用率时，使用 Softmax 值（即排除“不确定”标签）的简单阈值化提高了模型的性能，但没有提高对物种丰富度的估计。
4. 我们得出的结论是，通过在生态环境中进行充分的测试和评估，机器学习模型可以生成直接用于生态分析的标签，而无需手动验证。我们为用户社区提供了一个多平台、多语言的图形用户界面，可用于离线运行我们的模型。