Repository for Code and Dataset associated to the paper "Uncertainty Estimation for Data-driven Visual Odometry"
You can download the ISARLAB CAR dataset we gathered and used for the experiments in our publication from the following link:
http://sira.diei.unipg.it/supplementary/uavo_tro2020/ISARLAB_CAR.zip
It contains 15 sequences with associated absolute ground truth trajectory and intrinsic camera parameters. The i-th row in each GT file contains the 3x4 pose matrix associated to the i-th image of that sequence (each line in the file is built by concatenating the rows of the matrix). The intrisix.txt file contains the intrinsic matrix of the camera (as for the GT file, the row of the matrix are concatenated in a single line).
pytorch 3.8
pyquaternion
scipy
opencv-python
pyyaml
tensorboardX