Substantial progress has been made in the past 15 years regarding how prey use a variety of visual camouflage types to exploit both predator visual processing and cognition, including background matching, disruptive coloration, countershading and masquerade. By contrast, much less attention has been paid to how predators might overcome these defences. Such strategies include the evolution of more acute senses, the co-opting of other senses not targeted by camouflage, changes in cognition such as forming search images, and using behaviours that change the relationship between the cryptic individual and the environment or disturb prey and cause movement. Here, we evaluate the methods through which visual camouflage prevents detection and recognition, and discuss if and how predators might evolve, develop or learn counter-adaptations to overcome these. This article is part of the theme issue ‘Signal detection theory in recognition systems: from evolving models to experimental tests'.
Here we introduce BehaveAI, a biologically inspired video analysis framework that integrates static and motion information through a novel color-from-motion encoding strategy. This method translates object movement—direction, speed, and acceleration—into color gradients, enabling both human annotators and pre-trained convolutional neural networks (CNNs) to infer motion patterns while retaining high-resolution spatial detail. Using a range of case studies, we demonstrate how the increased salience of motion information allows for the robust detection of objects that are challenging or impossible to identify reliably from static frames alone, particularly in complex natural scenes. We further demonstrate the reliable classification of different behaviors in animals and single-celled organisms. Additionally, the framework supports flexible hierarchical model structures that can separate the tasks of detection and classification for optimal efficiency, and provide individual tracking data that specifies what is present where and what it is doing in each frame. The framework makes use of the latest deep learning architecture (YOLO11), combined with a semi-supervised annotation workflow. Together with salient motion information, these features can dramatically reduce the effort required for dataset annotation such that reliable models can often be made within an hour. Moreover, smaller annotation datasets mean that model training can be achieved on conventional computers without dedicated hardware, thereby improving accessibility. The motion encoding approach is also computationally lightweight, and can run in real-time on low-end edge devices such as a Raspberry Pi. We release the framework as a free, open source, and user-friendly package.
Here we introduce BehaveAI, a biologically inspired video analysis framework that integrates static and motion information through a novel colour-from-motion encoding strategy. This method translates object movement – direction, speed, and acceleration – into colour gradients, enabling both human annotators and pre-trained convolutional neural networks (CNNs) to infer motion patterns while retaining high-resolution spatial detail. Using a range of case studies, we demonstrate how the increased salience of motion information allows for the robust detection of objects that are challenging or impossible to identify reliably from static frames alone, particularly in complex natural scenes. We further demonstrate the reliable classification of different behaviours in animals and single-celled organisms. Additionally, the framework supports flexible hierarchical model structures that can separate the tasks of detection and classification for optimal efficiency, and provide individual tracking data that specifies what is present where and what it is doing in each frame. The framework makes use of the latest deep learning architecture (YOLO), combined with a semi-supervised annotation workflow. Together with salient motion information, these features can dramatically reduce the effort required for dataset annotation such that reliable models can often be made within an hour. Moreover, smaller annotation datasets mean that model training can be achieved on conventional computers without dedicated hardware, thereby improving accessibility. The motion encoding approach is also computationally lightweight, and can run in real-time on low-end edge devices such as a Raspberry Pi. We release the framework as a free, open source, and user-friendly package.
Nocturnal moths are severely affected by light pollution, most notoriously through fatal attraction to artificial lights, yet flight-to-light is not their only response. To investigate how artificial lights impact flight behaviour, we exposed over 1200 wild-caught moths of 62 species to LED lights with different characteristics, under varying background lighting conditions, and tracked over 500 flight paths in three dimensions. Flight-to-light behaviour and flight tortuosity both increased with light intensity, irrespective of spectrum, though tortuosity was affected by lower levels of white than amber light, suggesting white LEDs could impact moth trajectories from greater distances. Flight tortuosity was also higher upon exposure to a single light versus three producing equivalent illuminance. Conversely, higher background light levels led to reductions in both flight-to-light and tortuosity, but moths were also less likely to take flight in these conditions, suggesting that both point sources and diffuse background lighting disrupt moth movement. Finally, moths caught using light traps were less likely to fly and, if they did, more likely to fly towards light sources than those caught with butterfly nets. These findings suggest mitigation policies for light pollution should prioritize reducing light intensity, and point to new directions for future research.