未命名数据集

本研究开发了一个专门用于驾驶员疲劳检测的数据集，该数据集主要由两部分组成：第一部分是用于训练CNN的标记图像，第二部分是为LSTM网络开发的时间序列数据。数据集考虑了光照变化、驾驶员头部姿势等多种干扰因素，以提高神经网络训练的准确性和鲁棒性。通过使用PS3摄像头在模拟驾驶环境中录制视频，并从中提取了1042张图像。此外，数据集还包括了用于LSTM训练的二维时间序列数据，记录了10名参与者的睡眠阶段，以及眼睛闭合和打哈欠的持续时间。该数据集的应用领域主要集中在驾驶员疲劳检测和预测，旨在通过分析驾驶员的行为模式来提前预警潜在的疲劳驾驶风险。

This study presents a dedicated dataset for driver fatigue detection, which mainly comprises two parts: the first part consists of labeled images for CNN training, and the second part is time-series data tailored for LSTM networks. This dataset incorporates various interfering factors such as lighting variations and driver head poses, aiming to enhance the accuracy and robustness of neural network training. Videos were recorded in a simulated driving environment using a PS3 camera, from which 1,042 images were extracted. Additionally, the dataset includes 2D time-series data for LSTM training, which documents the sleep stages, eye closure durations and yawning durations of 10 participants. The primary application scenarios of this dataset focus on driver fatigue detection and prediction, with the objective of issuing early warnings for potential fatigued driving risks by analyzing drivers' behavioral patterns.

搜集汇总

背景与挑战

背景概述

该数据集是一个专门用于驾驶员疲劳检测的数据集，包含用于训练CNN的标记图像和用于LSTM的时间序列数据，考虑了光照变化和头部姿势等干扰因素以提高模型准确性。数据来源于模拟驾驶环境中录制的视频，包括1042张图像和10名参与者的睡眠阶段记录，主要应用于疲劳驾驶风险预警。