Generating Realistic 3D Models from SVO-Captured Point Clouds Using SDK


I am currently working on a project where I aim to collect 3D point clouds from a camera and then generate a 3D model. I believe this can be achieved by capturing video in SVO format and then processing it using the SDK. Could someone please confirm if my understanding is correct? If so, I would appreciate any advice or guidance on how to proceed with this approach.

Additionally, I have managed to obtain a depth map image from the camera and used it to generate point clouds. However, I’ve noticed that the resulting 3D model lacks realism, as it is derived from only a single viewpoint. Any suggestions for overcoming this limitation and achieving more realistic 3D models would be greatly appreciated.

Thank you!

Hi @code_lover,

You are correct, recording a SVO is a good way to record a sequence and run it afterward with the ZED SDK, where you can try our different depth and mapping parameters according to the results you are working for.

For generating 3D models, you can look into our spatial mapping module which fuses the information from multiple point clouds to produce either a fused point cloud, or a 3D mesh.

We also have a tool called ZEDfu which should help, it generates 3D models or point clouds from our spatial mapping module, and has an interface that can use either the live feed from your camera, or a recorded SVO file.


1 Like

Thank you very much, Matthew, for the helpful response.

Since I currently do not have access to the camera, could you please guide me on how to simply capture the SVO file? Additionally, if there are any other specific files or pieces of information I should request from my colleague who has access to the camera, I would greatly appreciate your advice on this. This will help me ensure that I have all the necessary data to work with the ZED SDK and ZEDfu tool effectively.

Thank you once again for your assistance!

Hi again,

Your colleague can use ZED_Explorer and press the record button at the bottom to film an SVO sequence. The only thing required is for him to send you the file, and you can import it in any of our tools, or use it in our code samples.

1 Like

Thank you so much for your assistance.

I have one final question regarding the capabilities of the ZED software. Is it possible to save the point clouds directly in PLY format and subsequently utilize Python libraries to work on them? If this functionality is available, could you please guide me on whether the ZED software allows for direct capturing and saving in PLY format, or should I use Python code for this process? Additionally, I am interested to know if the handling of point clouds in PLY format would be similar to the processing of point clouds generated from SVO files. Any guidance on these points would be highly valuable.

Thank you once again for all your help!

The ZEDfu tool saves fused point clouds as a PLY automatically if you set the “Spatial Map Type” to Fused Point Cloud in the settings.

If you wish to do this through code, you can use the Mesh::save method available in the API, which can also save to ply.