RIFIR – A Far Infrared Dataset

This dataset  consists of sequences acquired in an urban environment with two cameras (one Far Infrared and two color cameras in stereovision) mounted on the exterior of a vehicle.

Training dataset: 14788 frames (containing aprox. 19000 pedestrain bounding boxes in visible spectrum and aprox. 14000 in infrared spectrum) of 138 unique pedestrians

Testing dataset: 9373 frames  ( containing aprox. 7000 pedestrain bounding boxes in visible spectrum and aprox. 6000 in infrared spectrum) of 33 unique pedestrians

You can cite this dataset by:

Miron, Alina Dana. "Multi-modal, Multi-Domain Pedestrian Detection and Classification: Proposals and Explorations in Visible over StereoVision, FIR and SWIR." PhD diss., 2014.