LISA Traffic Light Dataset

## Context

When evaluating computer vision projects, training and test data are essential. The used data is a representation of a challenge a proposed system shall solve. It is desirable to have a large database with large variation representing the challenge, e.g detecting and recognizing traffic lights (TLs) in an urban environment. From surveying existing work it is clear that currently evaluation is limited primarily to small local datasets gathered by the authors themselves and not a public available dataset. The local datasets are often small in size and contain little variation. This makes it nearly impossible to compare the work and results from different author, but it also become hard to identify the current state of a field. In order to provide a common basis for future comparison of traffic light recognition (TLR) research, an extensive public database is collected based on footage from US roads. The database consists of continuous test and training video sequences, totaling 43,007 frames and 113,888 annotated traffic lights. The sequences are captured by a stereo camera mounted on the roof of a vehicle driving under both night- and daytime with varying light and weather conditions. Only the left camera view is used in this database, so the stereo feature is in the current state not used.

## Content

The database is collected in San Diego, California, USA. The database provides four day-time and two night-time sequences primarily used for testing, providing 23 minutes and 25 seconds of driving in Pacific Beach and La Jolla, San Diego. The stereo
image pairs are acquired using the Point Grey’s Bumblebee XB3 (BBX3-13S2C-60) which contains three lenses which capture images with a resolution of 1280 x 960, each with a Field of View(FoV) of 66°. Where the left camera view is used for all the test sequences and training clips. The training clips consists of 13 daytime clips and 5 nighttime clips.

### Annotations
The annotation.zip contains are two types of annotation present for each sequence and clip. The first annotation type contains information of the entire TL area and what state the TL is in. This annotation file is called frameAnnotationsBOX, and is generated from the second annotation file by enlarging all annotation larger than 4x4. The second one is annotation marking only the area of the traffic light which is lit and what state it is in. This second annotation file is called frameAnnotationsBULB.

The annotations are stored as 1 annotation per line with the addition of information such as class tag and file path to individual image files. With this structure the annotations are stored in a csv file, where the structure is exemplified in below listing:

``
Filename;Annotation tag;Upper left corner X;Upper left corner Y;Lower right corner X;Lower right corner Y;Origin file;Origin frame number;Origin track;Origin track frame number
``

## Acknowledgements
When using this dataset we would appreciate if you cite the following papers:

Jensen MB, Philipsen MP, Møgelmose A, Moeslund TB, Trivedi MM. [Vision for Looking at Traffic Lights: Issues, Survey, and Perspectives](http://ieeexplore.ieee.org/document/7398055/). I E E E Transactions on Intelligent Transportation Systems. 2016 Feb 3;17(7):1800-1815. Available from, DOI: 10.1109/TITS.2015.2509509

Philipsen, M. P., Jensen, M. B., Møgelmose, A., Moeslund, T. B., & Trivedi, M. M. (2015, September). [Traffic light detection: A learning algorithm and evaluations on challenging dataset](http://ieeexplore.ieee.org/document/7313470/). In intelligent transportation systems (ITSC), 2015 IEEE 18th international conference on (pp. 2341-2345). IEEE.

### Bibtex
```
@article{jensen2016vision,
title={Vision for looking at traffic lights: Issues, survey, and perspectives},
author={Jensen, Morten Born{\o} and Philipsen, Mark Philip and M{\o}gelmose, Andreas and Moeslund, Thomas Baltzer and Trivedi, Mohan Manubhai},
journal={IEEE Transactions on Intelligent Transportation Systems},
volume={17},
number={7},
pages={1800--1815},
year={2016},
doi={10.1109/TITS.2015.2509509},
publisher={IEEE}
}
```

```
@inproceedings{philipsen2015traffic,
title={Traffic light detection: A learning algorithm and evaluations on challenging dataset},
author={Philipsen, Mark Philip and Jensen, Morten Born{\o} and M{\o}gelmose, Andreas and Moeslund, Thomas B and Trivedi, Mohan M},
booktitle={intelligent transportation systems (ITSC), 2015 IEEE 18th international conference on},
pages={2341--2345},
year={2015},
organization={IEEE}
}
```

{'Context': '在评估计算机视觉项目时，训练和测试数据至关重要。所使用的数据应能体现拟解决挑战的实际情况。理想情况下，应拥有一套包含大量变体以反映挑战的大规模数据库，例如在城市环境中检测和识别交通信号灯（TL）。通过调查现有研究，明显看出当前评估主要局限于作者自行收集的小型本地数据集，而非公开可用的数据集。这些本地数据集通常规模较小，且变异度低。这几乎使得比较不同作者的工作和成果变得不可能，同时也难以识别某一领域的当前状态。为了为未来交通信号灯识别（TLR）研究提供共同的比较基础，我们基于美国道路的影像收集了一个广泛公开的数据库。该数据库由连续的测试和训练视频序列组成，总计43,007帧和113,888个标注的交通信号灯。这些序列由安装在驾驶车辆车顶的立体相机捕捉，该车在夜间和白天以不同的光照和天气条件下行驶。在此数据库中，仅使用左侧相机视角，因此立体特性目前尚未利用。', '### Annotations': 'annotation.zip文件包含每种序列和片段的两种类型的标注。第一种标注类型包含整个交通信号灯区域的信息以及交通信号灯的状态。此标注文件称为frameAnnotationsBOX，通过放大所有大于4x4的标注由第二个标注文件生成。第二种标注类型仅标记亮着的交通信号灯区域及其状态。此第二个标注文件称为frameAnnotationsBULB。

标注以每行一个标注的方式存储，并附加有关类别标签和单个图像文件路径等信息。以这种结构，标注存储在csv文件中，其结构如下所示：

``
Filename;Annotation tag;Upper left corner X;Upper left corner Y;Lower right corner X;Lower right corner Y;Origin file;Origin frame number;Origin track;Origin track frame number
``', 'Content': '该数据库在美国加利福尼亚州圣地亚哥收集。数据库提供了四个白天序列和两个夜间序列，主要用于测试，提供了23分钟25秒的太平洋海滩和拉霍亚的驾驶影像。使用Point Grey的Bumblebee XB3（BBX3-13S2C-60）获取立体图像对，该设备包含三个镜头，每个镜头以1280 x 960的分辨率捕捉图像，每个镜头的视场角（FoV）为66°。所有测试序列和训练片段均使用左侧相机视角。训练片段包括13个白天片段和5个夜间片段。', '## Acknowledgements': '在使用此数据集时，如能引用以下论文，我们将不胜感激：

Jensen MB，Philipsen MP，Møgelmose A，Moeslund TB，Trivedi MM. [Vision for Looking at Traffic Lights: Issues, Survey, and Perspectives](http://ieeexplore.ieee.org/document/7398055/). I E E E Transactions on Intelligent Transportation Systems. 2016 Feb 3;17(7):1800-1815. Available from, DOI: 10.1109/TITS.2015.2509509

Philipsen，M. P.，Jensen，M. B.，Møgelmose，A.，Moeslund，T. B.，& Trivedi，M. M. (2015, September). [Traffic light detection: A learning algorithm and evaluations on challenging dataset](http://ieeexplore.ieee.org/document/7313470/). In intelligent transportation systems (ITSC)，2015 IEEE 18th international conference on (pp. 2341-2345). IEEE。

### Bibtex

@article{jensen2016vision,
title={Vision for looking at traffic lights: Issues, survey, and perspectives},
author={Jensen, Morten Born{o} and Philipsen, Mark Philip and M{o}gelmose, Andreas and Moeslund, Thomas Baltzer and Trivedi, Mohan Manubhai},
journal={IEEE Transactions on Intelligent Transportation Systems},
volume={17},
number={7},
pages={1800--1815},
year={2016},
doi={10.1109/TITS.2015.2509509},
publisher={IEEE}
}

@inproceedings{philipsen2015traffic,
title={Traffic light detection: A learning algorithm and evaluations on challenging dataset},
author={Philipsen, Mark Philip and Jensen, Morten Born{o} and M{o}gelmose, Andreas and Moeslund, Thomas B and Trivedi, Mohan M},
booktitle={intelligent transportation systems (ITSC)，2015 IEEE 18th international conference on},
pages={2341--2345},
year={2015},
organization={IEEE}
}'}