Detection and tracking of belugas, kayaks and motorized boats in drone video using deep learning

Aerial imagery surveys are commonly used in marine mammal research to determine population size, distribution and habitat use. Analysis of aerial photos involves hours of manually identifying individuals present in each image and converting raw counts into useable biological statistics. Our research proposes the use of deep learning algorithms to increase the efficiency of the marine mammal research workflow. To test the feasibility of this proposal, the existing YOLOv4 convolutional neural network model was trained to detect belugas, kayaks and motorized boats in oblique drone imagery, collected from a stationary tethered system. Automated computer-based object detection achieved the following precision and recall, respectively, for each class: beluga = 74%/72%; boat = 97%/99%; and kayak = 96%/96%. We then tested the performance of computer vision tracking of belugas and manned watercraft in drone videos using the DeepSORT tracking algorithm, which achieved a multiple-object tracking accuracy (MOTA) ranging from 37% – 88% and multiple object tracking precision (MOTP) between 63% – 86%. Results from this research indicate that deep learning technology can detect and track features more consistently than human annotators, allowing for larger datasets to be processed within a fraction of the time while avoiding discrepancies introduced by labeling fatigue or multiple human annotators.

The Importance of Attention in the Twinkle Goes Effect

How do we perceive the location of moving objects? The position and motion literature is currently divided. Predictive accounts of object tracking propose that the position of moving objects is anticipated ahead of sensory signals, whilst non-predictive accounts claim that an anticipatory mechanism is not necessary. A novel illusion called the twinkle goes effect, describing a forward shift in the perceived final location of a moving object in the presence of dynamic noise, presents a novel opportunity to disambiguate these accounts. Across three experiments, we compared the predictions of predictive and non-predictive theories of object tracking by combining the twinkle goes paradigm with a multiple object tracking task. Specifically, we tested whether the size of the twinkle goes illusion would be smaller with greater attentional load (as entailed by the non-predictive, tracking continuation theory) or whether it would not be affected by attentional load (as entailed by predictive extrapolation theory). Our results failed to align with either of these theories of object localisation and tracking. Instead, we found evidence that the twinkle goes effect may be stronger with greater attentional load. We discuss whether this result may be a consequence of an essential, but previously unexplored relationship between the twinkle goes effect and representational momentum. In addition, this study was the first to reveal critical individual differences in the experience of the twinkle goes effect, and in the mislocalisation of moving objects. Together, our results continue to demonstrate the complexity of position and motion perception.

The Multiple Object Avoidance (MOA) task measures attention for action: Evidence from driving and sport

Performance in everyday tasks, such as driving and sport, requires allocation of attention to task-relevant information and the ability to inhibit task-irrelevant information. Yet there are individual differences in this attentional function ability. This research investigates a novel task for measuring attention for action, called the Multiple Object Avoidance task (MOA), in its relation to the everyday tasks of driving and sport. The aim in Study 1 was to explore the efficacy of the MOA task to predict simulated driving behaviour and hazard perception. Whilst also investigating its test–retest reliability and how it correlates to self-report driving measures. We found that superior performance in the MOA task predicted simulated driving performance in complex environments and was superior at predicting performance compared to the Useful Field of View task. We found a moderate test–retest reliability and a correlation between the attentional lapses subscale of the Driving Behaviour Questionnaire. Study 2 investigated the discriminative power of the MOA in sport by exploring performance differences in those that do and do not play sports. We also investigated if the MOA shared attentional elements with other measures of visual attention commonly attributed to sporting expertise: Multiple Object Tracking (MOT) and cognitive processing speed. We found that those that played sports exhibited superior MOA performance and found a positive relationship between MOA performance and Multiple Object Tracking performance and cognitive processing speed. Collectively, this research highlights the utility of the MOA when investigating visual attention in everyday contexts.