Observational studies of human behaviour often require the annotation of objects in video recordings. Automatic object detection has been facilitated strongly by the development of YOLO (‘you only look once’) and particularly by YOLOv8 from Ultralytics, which is easy to use. The present study examines the conditions required for accurate object detection with YOLOv8. The results show almost perfect object detection even when the model was trained on a small dataset (100 to 350 images). The detector, however, does not extrapolate well to the same object in other backgrounds. By training the detector on images from a variety of backgrounds, excellent object detection can be restored. YOLOv8 could be a game changer for behavioural research that requires object annotation in video recordings.
1 February 2014 in Journal of Experimental Psychology: Human Perception and Performance
Simultaneous masking refers to the impairment of performance on a visual target by simultaneously presented flankers. Whereas the spatial aspects of simultaneous masking have been studied extensively, the time course of these spatial influences is much less well understood. We here measure response latency and accuracy in a simultaneous masking paradigm and apply event history analysis to study the time course of target-flanker interactions. In our experiments, we presented a central target vernier flanked on both sides by 12 aligned distractor verniers that were either shorter, longer, or equal in length (Experiment 1), and that also were congruent or incongruent in their spatial offset with the target (Experiment 2). Response time distributions showed that there were more fast responses when the target was flanked by short flankers. Conditional accuracy functions showed that accuracy of responses dropped when the flankers had the same length as the target, but only for slow responses. These results are at odds with accounts based solely on lateral neural interactions or response competition, and instead suggest that top-down visual object-to-feature interference occurs when the target is not selected fast enough, congruent with object substitution theory.
Whether eye movements (as a measure of visual attention) contribute to the understanding of how multi-attribute decisions are made, is still a matter of debate. In this study, we show how machine learning methods can be used to separate the effects of the information presented, eye movement patterns, and attention to specific information. We also show how to deal with data from a relatively small sample of participants, often found in eye tracking studies that require in-lab testing. We make use of a dataset of 30 females who decided whether or not to accept screening for Chlamydia in 21 different scenarios. For this dataset, we find that eye movements did not add to the prediction of choice beyond the information presented to participants. Future studies should determine whether the same conclusion holds for other eye tracking datasets.
Eye movements have a spatial (where people look), but also a temporal (when people look) component. Various types of visualizations have been proposed that take this spatio-temporal nature of the data into account, but it is unclear how well each one can be interpreted and whether such interpretation depends on the question asked about the data or the nature of the dataset that is being visualised. In this study, four spatio-temporal visualization techniques for eye movements (chord diagram, scan path, scarf plot, space-time cube) were compared in a user study. Participants $$(N = 25)$$ answered three questions (what region first, what region most, which regions most between) about each visualization, which was based on two types of datasets (eye movements towards adverts, eye movements towards pairs of gambles). Accuracy of the answers depended on a combination of the dataset, the question that needed to answered, and the type of visualization. For most questions, the scan path, which did not use area of interest (AOI) information, resulted in lower accuracy than the other graphs. This suggests that AOIs improve the information conveyed by graphs. No effects of experience with reading graphs (for work or not for work) or education on accuracy of the answer was found. The results therefore suggest that there is no single best visualisation of the spatio-temporal aspects of eye movements. When visualising eye movement data, a user study may therefore be beneficial to determine the optimal visualization of the dataset and research question at hand. Graphical abstract
Negative attitudes toward offenders may hinder the rehabilitation process. The present study examines the relationship between attitudes toward sex offenders and stated acceptance of offenders and non‐offenders into various aspects of daily life. Sixty female members of the public (18–50 years old, UK residents, recruited by word of mouth and via social media) completed an attitudes towards sex offenders (ATS) scale and indicated for each of eight vignettes describing ex‐offenders and non‐offenders whether they would accept them in various situations (housing, employment, day‐to‐day activities). Results indicate that in this group of female participants, harsher attitudes toward sex offenders are associated with lower acceptance of sex offenders (around 50% less acceptance) and other offenders (around 25% less acceptance), but not non‐offenders, suggesting a tight coupling between attitudes and acceptance. The observed coupling between attitudes toward sex offenders and acceptance of offenders suggests that it will be difficult to change one without changing the other.
Gaze zone detection involves estimating where drivers look in terms of broad categories (e.g., left mirror, speedometer, rear mirror). We here specifically focus on the automatic annotation of gaze zones in the context of road safety research, where the system can be tuned to specific drivers and driving conditions, so that an easy to use but accurate system may be obtained. We show with an existing dataset of eye region crops (nine gaze zones) and two newly collected datasets (12 and 10 gaze zones) that image classification with YOLOv8, which has a simple command line interface, achieves near-perfect accuracy without any pre-processing of the images, as long as a model is trained on the driver and conditions for which annotation is required (such as whether the drivers wear glasses or sunglasses). We also present two apps to collect the training images and to train and apply the YOLOv8 models. Future research will need to explore how well the method extends to real driving conditions, which may be more variable and more difficult to annotate for ground truth labels.