Home Dataset Tutorials Active Vision Lab AVLCode GitHub
image test

RelocNet: Continuous Metric Learning Relocalisation Using Neural Nets

Below we provide some instructions on loading the data. If you need more help, emails us at relocnet@avlcode.org.

Images

The images are stored in the standard .ppm format and can be opened by most image libraries.

Depth

In RelocDB, the depth data is stored in the .exr format. We provide sample code to load the depth data in C++ and Python.

C++

The C++ loader of .exr files depends on the opencv library which can be installed using:

$ sudo apt-get install libopencv-dev
$ sudo apt-get install cmake
$ cd ./tutorials/cpp
$ mkdir build
$ cd build && cmake .. && make -j4
$ ./exr

The C++ code depends on OpenCV.

Python

The sample code for Python can be ran using:

$ sudo apt install openexr libopenexr-dev
$ pip install OpenEXR
$ python rgbd_display

The python code depends on Open3D and opencv-python.

Poses and Timestamps

We follow the TUM format for pose and timestamp representation.

Each line in the text file caputures one pose and corresponding timestamp, and is in the format timestamp tx ty tz qx qy qz qw, with: