An automated approach for counting groups of flying animals applied to one of the world's largest bat colonies
2023-06, Koger, Benjamin, Hurme, Edward, Costelloe, Blair R., O'Mara, M. Teague, Wikelski, Martin, Kays, Roland, Dechmann, Dina K. N.
Estimating animal populations is essential for conservation. Censusing large congregations is especially important since these are priorities for protection, but efficiently counting hundreds of thousands of moving animals remains a challenge. We developed a deep learning-based system using consumer cameras that not only counts but also records behavioral information for large numbers of flying animals in a range of lighting conditions including near darkness. We built a robust training set without human labeling by leveraging data augmentation and background subtraction. We demonstrate this approach by estimating the size of a straw-colored fruit bat (Eidolon helvum) colony in Kasanka National Park, Zambia with cameras encircling the colony to record evening emergence. Detection of bats was robust to deteriorating lighting conditions and changing backgrounds. Combined over five days, our population estimates ranged between 750,000 and 976,000 bats with a mean of 857,233. In addition to counts, we extracted wingbeat frequency, flight altitude, and local group polarity for 639,414 individuals. This open access method is an inexpensive but powerful approach that, in addition to radial emergences from central locations, can also be applied to unidirectional movements of flying groups, such as migratory streams of birds.
Using computer vision to study animal behavior in natural environments
2022, Koger, Benjamin
Understanding how animals behave in the context of their physical and social environments is fundamental to the study of animal behavior. While observing animal behavior in the field is always challenging, it becomes much harder, and in some cases even impossible when multiple animals are behaving together. Despite this challenge, collective dynamics are of wide interest across fields ranging from robotics to neuroscience to conservation. While catching and tagging animals is a common way to study animal movement dynamics in complex environments, this approach has many drawbacks. As an alternative, I highlight the utility of pairing high-resolution imaging with convolutional neural networks to study behavior in the field. In this thesis, I present three chapters, each of which presents a new method that allows researchers to extract high quality data about groups of animals behaving in naturalistic environments.
In chapter 1, I demonstrate a system that, using ten inexpensive consumer cameras, can robustly estimate the population size of the largest fruit bat colony in Africa. Beyond population estimates, this method also quantifies behavioral measures like flight altitude, wing beat frequency and local neighborhood polarity for thousands of bats each night. I combine threshold-based object detection techniques with neural networks to achieve the robustness of neural networks but without humans having to explicitly annotate training data.
In chapter 2, I demonstrate a system that combines imaging drones, machine learning, and landscape mapping techniques to record the movement of free ranging groups of animals at high spatial and temporal resolution across natural landscapes. This approach reconstructs the trajectories of all individuals in the group in 3D landscape maps generated from the same 4 observation videos while also quantifying individual features like animal posture, species, and age/sex class.
Finally, in chapter 3, I demonstrate a system that uses semantic segmentation to classify, for every pixel in an image of a honey bee frame, both its comb cell type as well as any contents that it may contain (honey or pollen for example). For the first time, this purely automated method allows researchers to study, through data about the nest, how honey bees, a model organism for distributed organization, build their nests and allocate space within it. Combined, these three chapters clearly demonstrate the potential for using automated image-based approaches to study animal behavior in complex natural environments.
Quantifying the movement, behaviour and environmental context of group‐living animals using drones and computer vision
2023-03-21, Koger, Benjamin, Deshpande, Adwait, Kerby, Jeffrey T., Graving, Jacob M., Costelloe, Blair R., Couzin, Iain D.
1. Methods for collecting animal behaviour data in natural environments, such as direct observation and biologging, are typically limited in spatiotemporal resolution, the number of animals that can be observed and information about animals'social and physical environments. 2. Video imagery can capture rich information about animals and their environments, but image-based approaches are often impractical due to the challenges of processing large and complex multi-image datasets and transforming resulting data, such as animals' locations, into geographical coordinates. 3. We demonstrate a new system for studying behaviour in the wild that uses drone-recorded videos and computer vision approaches to automatically track the location and body posture of free-roaming animals in georeferenced coordinates with high spatiotemporal resolution embedded in contemporaneous 3D landscape models of the surrounding area. 4. We provide two worked examples in which we apply this approach to videos of gelada monkeys and multiple species of group-living African ungulates. We demonstrate how to track multiple animals simultaneously, classify individuals by species and age–sex class, estimate individuals' body postures (poses) and extract environmental features, including topography of the landscape and animal trails. 5. By quantifying animal movement and posture while reconstructing a detailed 3D model of the landscape, our approach opens the door to studying the sensory ecology and decision-making of animals within their natural physical and social environments.
DeepPoseKit, a software toolkit for fast and robust animal pose estimation using deep learning
2019-10-01, Graving, Jacob M., Chae, Daniel, Naik, Hemal, Li, Liang, Koger, Benjamin, Costelloe, Blair R., Couzin, Iain D.
Quantitative behavioral measurements are important for answering questions across scientific disciplines-from neuroscience to ecology. State-of-the-art deep-learning methods offer major advances in data quality and detail by allowing researchers to automatically estimate locations of an animal's body parts directly from images or videos. However, currently-available animal pose estimation methods have limitations in speed and robustness. Here we introduce a new easy-to-use software toolkit, DeepPoseKit, that addresses these problems using an efficient multi-scale deep-learning model, called Stacked DenseNet, and a fast GPU-based peak-detection algorithm for estimating keypoint locations with subpixel precision. These advances improve processing speed >2× with no loss in accuracy compared to currently-available methods. We demonstrate the versatility of our methods with multiple challenging animal pose estimation tasks in laboratory and field settings-including groups of interacting individuals. Our work reduces barriers to using advanced tools for measuring behavior and has broad applicability across the behavioral sciences.
Manipulating nest architecture reveals three-dimensional building strategies and colony resilience in honeybees
2023, Marting, Peter R., Koger, Benjamin, Smith, Michael L.
Form follows function throughout the development of an organism. This principle should apply beyond the organism to the nests they build, but empirical studies are lacking. Honeybees provide a uniquely suited system to study nest form and function throughout development because we can image the three-dimensional structure repeatedly and non-destructively. Here, we tracked nest-wide comb growth in six colonies over 45 days (control colonies) and found that colonies have a stereotypical process of development that maintains a spheroid nest shape. To experimentally test if nest structure is important for colony function, we shuffled the nests of an additional six colonies, weekly rearranging the comb positions and orientations (shuffled colonies). Surprisingly, we found no differences between control and shuffled colonies in multiple colony performance metrics—worker population, comb area, hive weight and nest temperature. However, using predictive modelling to examine how workers allocate comb to expand their nests, we show that shuffled colonies compensate for these disruptions by accounting for the three-dimensional structure to reconnect their nest. This suggests that nest architecture is more flexible than previously thought, and that superorganisms have mechanisms to compensate for drastic architectural perturbations and maintain colony function.