WOLO: Wilson Only Looks Once – Datasets

WOLO: Wilson Only Looks Once – Estimating ant body mass from reference-free images using deep convolutional neural networks Abstract Size estimation is a hard computer vision problem with widespread applications in quality control in manufacturing and processing plants, livestock management, and studies on animal behaviour. Typically, image-based size estimation is facilitated by either well-controlled imaging conditions, the provision of global cues, or both. Reference-free size estimation is challenging, because objects of vastly different sizes can appear identical if they are of similar shape. Here, we attempt to implement automated and reference-free body size estimation to facilitate large-scale experimental work in a key model species in sociobiology: the leaf-cutter ants. Leaf-cutter ants are a suitable testbed for reference-free size-estimation, because their workers differ vastly in both size and shape; in principle, it is therefore possible to infer body mass, a proxy for size, from relative body proportions alone. Inspired by earlier work by E.O. Wilson, who trained himself to discern ant worker size from visual cues alone, we used various deep learning techniques to achieve the same feat automatically, quickly, and at scale from a single reference image: Wilson Only Looks Once (WOLO). Utilizing over 3 million hand-annotated and computer-generated images, a set of deep neural networks---including regressors, classifiers, and detectors---were trained to estimate the body mass of ants from image cut-outs. The WOLO networks approximately matched human performance, measured for a small group of both experts and non-experts, but were about 1000 times faster. Further refinement may enable accurate, high-throughput, and non-intrusive body mass estimation at scale, and so eventually contribute to a more nuanced and comprehensive understanding of the complex division of labour that characterises polymorphic insect societies. Datasets This repository contains all (cropped) frame datasets for training and benchmarking. Refer to the WOLO GitHub page for usage and the manuscript for dataset details. As the full-frame datasets are too large to host online permanently, please get in touch with the lead author (Fabian Plum) if you require access.WOLO：威尔逊一眼识蚁——基于深度卷积神经网络（deep convolutional neural networks）的无参考图像蚂蚁体重估测 摘要 尺寸估算是计算机视觉领域的一项挑战性问题，在制造与加工厂质检、畜牧管理以及动物行为研究中均具有广泛应用场景。通常，基于图像的尺寸估算可通过以下任一方式或二者结合实现：严格控制成像环境，或提供全局参考标识。无参考尺寸估算则极具挑战性，因为尺寸差异极大的物体若外形相似，在图像中会呈现出完全一致的视觉效果。本研究旨在实现自动化无参考体型估算，以助力社会生物学领域关键模式物种——切叶蚁的大规模实验研究。切叶蚁是无参考尺寸估算的理想测试对象：其工蚁在体型与外形上均存在显著差异，因此理论上仅通过身体相对比例即可推断作为体型替代指标的体重。受E.O.威尔逊早期研究启发——威尔逊通过自主训练，仅依靠视觉线索即可分辨蚁工蚁体型——我们采用多种深度学习技术，仅通过单张参考图像即可自动、快速且大规模地实现同款识别能力，该方法命名为“威尔逊一眼识蚁（Wilson Only Looks Once，WOLO）”。我们借助超过300万张人工标注与计算机生成的图像，训练了一组深度神经网络（deep neural networks）——包含回归器、分类器与目标检测器——以从蚂蚁图像裁剪区域中估算其体重。经针对少量专家与非专家群体的测试，WOLO模型的估算性能大致与人类相当，但速度提升了约1000倍。后续进一步优化后，该方法可实现大规模、高精度、高通量且非侵入式的体重估算，最终助力我们更细致全面地理解多态昆虫社会所特有的复杂劳动分工机制。 数据集 本仓库包含所有用于训练与基准测试的（裁剪后）帧数据集。使用方法请参阅WOLO的GitHub页面，数据集细节请参见论文手稿。 由于完整帧数据集体积过大，无法长期在线托管，若您需要获取数据集权限，请联系第一作者Fabian Plum。