- 2021-07-12: Python bindings now available here.
- 2020-05-06: Added Pangolin for visualization
- CMake >= 3.0
- OpenCV >= 3.4
- Eigen3 >= 3.3
- CUDA >= 10.0 (Although older versions might work)
- Pangolin (Optional, only for visualization)
To run the example, make sure you have Pangolin installed. The example should be automatically built during compilation.
DatasetLoader
looks for three files underneath the root path of the dataset:
calibration.txt
contains all neccessary intrinsic parameters needed for backprojecting depth. It shoud be a one liner with the format offx fy cx cy
, separated by space.associated.txt
contains the list of color and depth images including their time stamps.groundtruth.txt
have the ground truth poses for all frames. This usually comes with some datasets but you could also provide your own (by running a SLAM system for example).
Run the example with the following format:
./PangolinDisplay3D <path-to-dataset>/
This work is derived from VoxelHashing and InfiniTAM