I want to use TUM kinect datasets with ground truths which were published which can be found here. http://vision.in.tum.de/data/datasets/rgbd-dataset/download

I have a visual odometry code (using ICP) which needs both depth image and pointcloud, and found that the published datasets dont have them packed into one file. I found a script file addpointcloudsto_bagfile.py which adds pointclouds to ROSBAG file.

1. Should I save the script file in .sh format or .py format ?
2. When I run it in terminal, I am getting an error as below. I have cvbridge in the ROSPACKAGEPATH because roscd cvbridge takes me to its location

./addpointcloudstobagfiles.sh rgbddatasetfreiburg1desk2.bag Traceback (most recent call last): File "./addpointcloudstobagfiles.sh", line 32, in from cvbridge import CvBridge, CvBridgeError ImportError: No module named cv_bridge

3.The visual odometry code gives the position and orientation taking the first pose as origin, but the datasets have some offset. So How can this be adjusted ?

4.The axes of the frames which the datasets publish and the output which i get from visual odometry code may not be aligned, so how can this be adjusted?

The visual odometry code which I have basically gives an output as tf(estimated from ICP).