原文链接：https://arxiv.org/abs/1902.10859v1
代码：https://github.com/polarisZhao/PFLD-pytorch
论文名称
*
– Drowsy Driver Detection Using Two Stage Convolutional Neural Networks
– Visual-based Real Time Driver Drowsiness Detection System Using CNN
– Driver Drowsiness Detection Based on Joint Monitoring of Yawning, Blinking and Nodding
– IET Image Processing – 2021 – Zhao – Research on fatigue detection based on visual features
– Multi-Timescale Drowsiness Characterization Based on a Video of a Driver’s Face
– Federated Learning for Driver Status Monitoring
– temporal and spatial feature based approaches in drowsiness detection using deep learning technique
Drowsy Driver Detection Using Two Stage Convolutional Neural Networks
- Face Detection uses YOLO-V3
- Drowsy Classification
- 使用Inception-v3网络对闭眼数据训练，认为闭眼就是疲劳。对疲劳分类就定义为对睁眼和闭眼分类。
- 数据集（包括睁眼和闭眼图像）：CEW；提取Nth-DDD数据集中一些视频帧作为一个数据集；自己创建的数据集；
- 准确率80.32%, 79.34%and 89.90%
Visual-based Real Time Driver Drowsiness Detection System Using CNN
- Face Detection Viola&Jones算法
- Drowsy Classification
- 疲劳定义：自己设计浅层cnn网络判断是否闭眼，如果”闭眼”持续超过4个连续帧，系统将确定驾驶员处于睡眠状态
- 数据集：我们将Nth-DDD数据集视频序列的每一帧标记为”闭上眼睛”或”睁开眼睛”。总共有4620张人脸图像被标记，其中2310张是”闭着眼睛”的，2310张是”睁开眼睛”的。 准确率98.95%
Driver Drowsiness Detection Based on Joint Monitoring of Yawning, Blinking and Nodding
- Face Detection Viola&Jones算法
- Drowsy Classification
- 疲劳定义：根据人脸关键点算法算出，EAR，MAR，头部动向三个特征，将三个特征放在MLP，K近邻算法中做各种疲劳行为的分类，比如疲劳，闭眼，打哈欠，点头。
IET Image Processing – 2021 – Zhao – Research on fatigue detection based on visual features
- 这篇没啥好看的
Multi-Timescale Drowsiness Characterization Based on a Video of a Driver’s Face
- Face Detection OpenCV implementation of the Viola & Jones algorithm
- Drowsy Classification 使用CNN得到眼睑闭合度，收集眼睑闭合度的一段时间序列，对此时间序列进行疲劳分类训练。
Federated Learning for Driver Status Monitoring
- Face Detection
- Drowsy Classification 得到眼睛是否闭眼，嘴巴是否张开，然后结合PERCLOS准则判断疲劳，但是没有说具体·使用了多少帧，阈值是多少，与现在我们的实现方法是一样的。
temporal and spatial feature based approaches in drowsiness detection using deep learning technique
- Face Detection
- Drowsy Classification
根据人脸关键点算法算出，EAR，MAR，瞳孔位置等特征，将三个特征的时间序列放在LSTM（RNN的一种·）中做疲劳的分类
Original: https://blog.csdn.net/Icandescence_/article/details/123619552
Author: Icandescence_
Title: 疲劳检测方法总结_计算机视觉
原创文章受到原创版权保护。转载请注明出处：https://www.johngo689.com/532681/
转载文章受原作者版权保护。转载请注明原作者出处！