Overview

The SCUT FIR Pedestrian Datasets is a large far infrared pedestrian detection dataset. It consist of about 11 hours-long image sequences ($\sim 10^6​$ frames) at a rate of 25 Hz by driving through diverse traffic scenarios at a speed less than 80 km/h.

Data Collection

The image sequences were collected from 11 road sections under 4 kinds of scenes including downtown, suburbs, expressway and campus in Guangzhou, China.

Data Annotation

We annotated 211,011 frames for a total number of 477,907 bounding boxes around 7,659 unique pedestrians.

Instruction

• Seq video format. Data Format is compatible with Caltech Pedestrian Dataset Format
• datatool. Evaluation/labeling code for our dataset which is based on Caltech Dataset.
• toolbox. The datatool depended tool which is based on Piotr's Matlab Toolbox.
• pydatatool. If you want to use this dataset as coco style annotation in Detectron framework, please use this python version datatool.

citation

If you find SCUT FIR Pedestrian Dataset useful in your research, please consider citing:

@article{xu2019,
author = {Xu, Zhewei and Zhuang, Jiajun and Liu, Qiong and Zhou, Jingkai and Peng, Shaowu},
title = {{Benchmarking a large-scale FIR dataset for on-road pedestrian detection}},
journal = {Infrared Physics {\&} Technology},
pages = {199--208},
volume = {96},
year = {2019}
doi = {https://doi.org/10.1016/j.infrared.2018.11.007},
issn = {1350-4495},
}

License

BSD-2-Clause