Data Collection - ksu-mechatronics-research/deep-visual-odometry GitHub Wiki

To collect your own data, you need to set up the necessary visual pipeline. This involves getting the camera intrinsic calibration matrix and correcting the image. Then you need to label the data using a traditional visual odometry method. We will use ORB_SLAM2 or rpg_svo

Setting up a raspberry pi for collection:

Follow this tutorial to get started: https://gist.github.com/nickoala/8d7e0bc24be3cec459e63bc1eb8cc858

Then install ROS and anything else needed to gather data.

We are planning on using usb_cam to capture the video in ros and send it to the visual odometry program for labeling. We will find out if the pi has enough power to do this in real time.

Getting Camera Calibration Data:

Follow the tutorial at:

http://wiki.ros.org/camera_calibration

Then collect your data

Use opencv or some other method to record the video you are using.