Side Profile Vehicle Binary Images

This is a dataset of vehicle's binary images obtained with an optical barrier.
The main goal off this dataset is to train and test a vehicle classification method.
Categories are defined according to the number of axles and the presence of double wheel or not. (See the vehicle_categories.png)

This dataset is composed by 3 folders.
Train: Images used to train a CNN. (Over 1000 images/cat) Total of 11232 images.
Experiment1: Images used to validate.5389 images.
Experiment2: Field test. Four days off images. 196018 images.
Train and Experiment1 are selected images from the same lane, in different periods over 3 months.
Experiment2 has images from several different lanes, and 3 different toll stations. We do not delete any image in this set.

本数据集为通过光学挡板（optical barrier）采集得到的车辆二值图像数据集。本数据集的核心目标为训练并测试车辆分类方法。类别依据车轴数量以及是否配备双轮结构进行定义（详见vehicle_categories.png）。
本数据集包含3个文件夹。训练集（Train）：用于训练卷积神经网络（Convolutional Neural Network, CNN）的图像，每类图像数量超过1000张，总样本量为11232张。实验集1（Experiment1）：用于模型验证的图像，共计5389张。实验集2（Experiment2）：现场测试数据集，采集自4天的图像数据，总样本量为196018张。
训练集与实验集1的图像均选自同一条车道，采集时段覆盖3个月以上的不同时期。实验集2的图像取自多条不同车道以及3个不同收费站，该数据集未对任何图像进行删除处理。