This is a vehicle dataset for vision-based place recognition with manually ground truthed frame correspondences. The dataset was captured in two different conditions for the same route: one on a sunny day and one during a rainy night. Both datasets were taken in the suburb of Alderley, Queensland.
|Example Frame Match from Dataset|
|Night Traverse (Image03670.jpg)||Day Traverse (Image02477.jpg)|
Attached below are the original .avi movie files as well as the individual frames, extracted from the video and compressed into zip files. A .csv file holds the frame correspondences, which is replicated in the MATLAB mat file (requires MATLAB to use). The frame correspondences can be visually validated using the MATLAB script PLAYBACK.m. This will play back the videos, aligned using the frame correspondences in fm.mat.
FRAMESA corresponds to the night time traverse and FRAMESB to the day time traverse.
|Original Video Files (.avi format)||Extracted Frames||Frame Correspondences||MATLAB Resources|
FRAMESA.zip - this contains the frames from night1_orig.avi
FRAMESB.zip - this contains the frames from day1_orig.avi
|framematches.csv - column 1 is FRAMESA and column 2 is FRAMESB|
fm.mat - frame correspondences
PLAYBACK.m - for validation of ground truth
Other Papers Using this Dataset
M Milford, W Scheirer, E Vig, A Glover, O Baumann, J Mattingley, D Cox, "Condition-Invariant, Top-Down Visual Place Recognition", in IEEE International Conference on Robotics and Automation, Hong Kong, China, 2014.