DriveFace 司机驾驶场景时的面部特征数据集

DriveFace 数据库包含在真实场景中驾驶时的对象图像序列。它由 606 个样本组成，每个样本 640 × 480 像素，在不同的日子里从 4 名司机（2 名女性和 2 名男性）获得，这些司机有几个面部特征，如眼镜和胡须。 附加文件：drivFace.mat ，该包含 Matlab 中的数据集（在 prtools 库下），驾驶员面部标准化为每个 80x80 像素，其相关的注视方向标签为“向右看”、“正面”和“向左看”。 Attribute Information: 地面实况包含人脸边界框的注释和人脸关键点（眼睛、鼻子和嘴巴）。 给出了一组标签，将每个图像分配到 3 个可能的注视方向类中。 第一个类是“向右看”类，包含 -45º 和 -30º 之间的头部角度。 第二个是“frontal”类，包含 -15º 和 15º 之间的头部角度。 最后一个是“向左看”类，包含 30º 和 45º 之间的头部角度。 文件和脚本 • DriveImages.zip 包含驱动程序图像。图像名称的格式为：* YearMonthDay_subject_Driv_imNum_HeadPose.jpg即 20130529_01_Driv_011_f .jpg 是 11 序列图像对应的 fisrts driver 的帧，头部姿势是正面。主题 = [1:4]，imNum = [001:...]，HeadPose = lr（向右看），f（正面）和 lf（向左看）。 drivPoints.txt包含表格格式的地面真相，其中各列有以下信息。 * fileName是DrivImages.zip中的图像名称。 * subject = [1:4］ * imgNum = int * label = [1/2/3]（头部姿势类，分别对应于[lr/f/lf]，) * ang = [-45, -30/ -15 0 15/ 30 15] (头部姿势角度) * [xF yF wF hF] =面部位置 * [xRE yRE] = 严格的眼睛位置 * [xLE yL] = 左眼位置 * [xN yN] = 鼻子的位置 * [xRM yRM] = 严格的嘴角位置 * [xLM yLM] = 嘴的左角 read_drivPoints.m是一个Matlab函数，用于读取drivPoints文件。你也可以使用。* Table = readtable('drivPoints.txt')。 drivFace.mat包含了Matlab中的数据集（在prtools库下），其中的驾驶员面孔被归一化为80x80像素，以及他们相关的凝视方向标签 "向右看"、"正面 "和 "向左看"。 Citation Request: Katerine Diaz-Chito, Aura Hernández-Sabaté, Antonio M. López, A reduced feature set for driver head pose estimation, Applied Soft Computing, Volume 45, August 2016, Pages 98-107, ISSN 1568-4946,

The DriveFace database contains image sequences of drivers captured while driving in real-world scenarios. It consists of 606 samples, each with a resolution of 640 × 480 pixels, collected from 4 drivers (2 female and 2 male) across different days. These drivers have various facial features such as glasses and beards.

Additional file: drivFace.mat, which contains the dataset in Matlab (under the prtools library), where driver faces are normalized to 80×80 pixels, with their associated gaze direction labels: "looking right", "frontal", and "looking left".

Attribute Information:
The ground truth includes annotations for facial bounding boxes and facial landmarks (eyes, nose, and mouth). A set of labels is provided to assign each image to one of 3 possible gaze direction classes. The first class is the "looking right" class, covering head angles between -45° and -30°. The second is the "frontal" class, covering head angles between -15° and 15°. The last is the "looking left" class, covering head angles between 30° and 45°.

Files and Scripts
• DriveImages.zip contains the driver images. The naming format of the images is: *YearMonthDay_subject_Driv_imNum_HeadPose.jpg. For example, 20130529_01_Driv_011_f.jpg is the 11th frame of the first driver’s sequence, with a frontal head pose. Here, subject ∈ [1:4], imNum ∈ [001:...], and HeadPose corresponds to lr (looking right), f (frontal), and lf (looking left).

drivPoints.txt contains ground truth in tabular format, with the following information per column:
* fileName: the image name within DrivImages.zip
* subject ∈ [1:4]
* imgNum: integer
* label ∈ [1/2/3] (head pose classes, corresponding to [lr/f/lf] respectively)
* ang ∈ [-45, -30 / -15 0 15 / 30 15] (head pose angles)
* [xF yF wF hF]: facial bounding box position
* [xRE yRE]: position of the right eye
* [xLE yL]: position of the left eye
* [xN yN]: position of the nose
* [xRM yRM]: position of the right mouth corner
* [xLM yLM]: position of the left mouth corner

read_drivPoints.m is a Matlab function for reading the drivPoints file. You can also use: Table = readtable('drivPoints.txt').

drivFace.mat contains the dataset in Matlab (under the prtools library), where driver faces are normalized to 80×80 pixels, along with their associated gaze direction labels "looking right", "frontal", and "looking left".

Citation Request:
Katerine Diaz-Chito, Aura Hernández-Sabaté, Antonio M. López, A reduced feature set for driver head pose estimation, Applied Soft Computing, Volume 45, August 2016, Pages 98-107, ISSN 1568-4946.

搜集汇总

背景与挑战

背景概述

DriveFace数据集是一个包含606个驾驶场景面部图像的数据集，图像分辨率为640×480像素，采集自4名司机。数据集提供了详细的面部特征标注和注视方向标签，适用于驾驶行为分析和面部识别研究。