RGB-P Car Detection Dataset

Wen Dong, Haiyang Mei, Ziqi Wei, Ao Jin, Sen Qiu, Qiang Zhang, Xin Yang.

Exploiting Polarized Material Cues for Robust Car Detection.In Proceedings of the AAAI Conference on Artificial Intelligence. 2024.

https://arxiv.org/abs/2401.02606

Introduction:

RGBP-Car  is the first pixel-aligned RGB-polarization car detection dataset with trichromatic polarization 

measurements. We record cars in diverse traffic scenes using FLIR-Blackfly-S, a polarized color camera 

that simultaneously obtains pixel-aligned polarization measurements in four linear polarization directions 

(0◦, 45◦, 90◦,  and 135◦) for each color channel (i.e. , R, G, and B). RGBP-Car contains 2601 RGB, AoLP, 

and  DoLP image triplets. Each image has manually  labeled bounding boxes indicating the position and 

size of each car.  To ensure the diversity and challenge of our dataset, we take the RGB-P images under 

different weather conditions (clear  and  rainy), different lighting conditions (daytime  and  nighttime), diff-

erent driving environments (indoor, outdoor, road and parking lot),and different car densities.