tzutalin / OpenCV-RgbdOdometry

:snowman: OpenCV RGBD-Odometry (Visual Odometry based RGB-D images)

RGBD-Odometry (Visual Odometry based RGB-D images)

Requirement

OpenCV >= 3.0

Setup

Download RGB-D dataset from TUM or KTTI

main.cpp will read the path of color and depth images from ./assoc.txt, then run OpenCV RgbdOdometry to compute visual odometry

The format of assoc.txt looks like:

timestamp1 rgb/[color_image_filename1] timestamp1 depth/[depth_image_filename1]
...
timestampN rgb/[color_image_filenameN] timestampN depth/[depth_image_filenameN]

You should change camera paramerts at the top of main.cpp

#define FOCUS_LENGTH 525.0
#define CX 319.5
#define CY 239.5

Build & Run

For the first time, you should download the dataset. You can use the below command

$ python tools/download_dataset.py

Start building

$ mkdir -p build; cd build

$ cmake ..; make

Start running

$ cd [Opencv-RgbdOdometry]

Create assoc.txt having synchronized rgb and depth images

$ cd rgbd_dataset_freiburg2_pioneer_slam3

$ python associate.py rgb.txt depth.txt > assoc.txt

$ ../build/rgbd-odometry

Demo video