These datasets were created by Rui (Ranger) Fan.
The datasets were used for stereo vision-based road surface 3D reconstruction and pothole detection. Kindly refer to our paper: "Road Surface 3D Reconstruction Based on Dense Subpixel Disparity Map Estimation", published in IEEE Transactions on Image Processing (T-IP), for more details.
In this repository, there are three datasets (dataset 1, dataset 2, and dataset 3).
Datasets 1 and 2 mainly aim at road surface 3D recontruction, while dataset 3 is used to evaluate the road surface 3D reconstruction accuracy.
Each dataset includes uncalibrated left and right images and calibrated left and right images. The calibrated images are used for disparity map estimation, while the uncalibrated images and calibration parameters are used to reconstruct the 3D road geometry.
A demo video can be found on YouTube.
e-mail: rui.fan@ieee.org