Is there any way to use Point Cloud Library (or similar) to get a skeleton from point cloud data like obj?

Question

everyone.

I've been stuck for some days searching some way to get the skeleton of a point cloud data (like OBJ) but not using kinect. Is it possible? I found the Point Cloud Library which does a lot of tasks related to point cloud data, and in their documentation there is a body keypoints detector, but it also works with kinect grabbers. In my case, I have a point cloud data like in the picture, which was generated by another depth sensor scanner. Is it possible to find the key points in such data?

I really would appreciate any help. Thanks in advance.

Answer 1

Even if it's not explicitly mentioned in the tutorial you linked, a quick to the code suggests that you can use different data sources (e.g. PCD files), so you're not stuck with the live capture from Kinect.

All the tutorial code really does is the following:

• Setup the GPU for the people parts detection.
• Pick the appropriate data source.
• Load the tree files for the body part detector.
• Run the PeopleDetector on a single frame captured from the live grabber stream/PCD file.