Annotations directory

<b>Directory content</b>: This directory contains 21 .CSV files reporting the data about 23 specific drivers’ mannerisms and behaviors (e.g., rubbing/holding face, yawning) observed during the driving session (ID_Number_LabelData).<b>Method and instruments</b>: To collect the annotation data, two trained and independent raters employed a customized video analysis tool (HADRIAN’s EYE software; Di Stasi et al., 2023) to identify and annotate drivers’ fatigue- and sleepiness-related mannerisms and behaviors. The tool allows the synchronized reproduction of the videos obtained through the RGB camera (for further details, see RGB and Depth videos directory) and the recording of the main central screen of the simulator (for further details, see Driving simulator indices directory). Both videos were automatically divided by the HADRIAN’s EYE software into a series of 5-min 39 chunks that were then shuffled and presented to the raters in a randomized order to minimize bias (e.g., overestimating the level of fatigue towards the end of the experimental session). Then, a customized Matlab code (Mathworks Inc., Natick, MA, USA) was used to detect discrepancies in the outputs of the two raters. Two types of discrepancies were detected: (i) type, and (ii) timing of the detected mannerism/behavior. Finally, in case of discrepancies, a third independent rater reviewed the videos to solve the issue.<b>目录内容</b>：本目录包含21个逗号分隔值（CSV）文件，记录了23名驾驶员在驾驶过程中被观测到的特定动作与行为数据（例如搓脸/托脸、打哈欠），文件命名格式为ID_Number_LabelData。<b>研究方法与仪器设备</b>：为采集标注数据，两名经过培训的独立评分员采用定制化视频分析工具（HADRIAN’s EYE软件；Di Stasi等，2023），对驾驶员与疲劳、嗜睡相关的动作及行为进行识别与标注。该工具可同步回放通过RGB相机采集的视频（详细信息参见RGB与深度视频目录），并可记录驾驶模拟器的主中央屏幕画面（详细信息参见驾驶模拟器指标目录）。该软件会自动将两段视频分割为若干个时长为5分39秒的片段，随后对这些片段进行随机打乱，并以随机顺序呈现给评分员，以尽可能降低偏倚（例如避免因实验临近结束而高估疲劳程度）。随后，采用定制化Matlab代码（美国马萨诸塞州内蒂克市Mathworks公司开发）对两名评分员的标注结果进行差异检测。本次检测共发现两类差异：一是（i）所识别动作/行为的类别差异，二是（ii）该动作/行为的时间节点差异。最终，若出现评分差异，则由第三名独立评分员重新审阅视频以解决分歧。