The example shows how to run stereo visual SLAM using recorded data. It doesn't support "online" visual SLAM yet, meaning that you need to use your stereo camera to record a video, save it to your local drive, and then run the algorithm using the video.
Once you save the video to your local disk, you can read each frame using VideoReader:
reader = VideoReader('myvideo.avi');
currI = read(reader, currFrameIdx);
If this off-line workflow is what you are trying to do, then you can start with stereo camera calibration using the Stereo Camera Calibrator app. Once you get the intrinsic parameters of the camera, you can easily follow the steps in the example to run the stereo visual SLAM pipeline.