Huge scene reconstruction

The RealityCapture application (RC) uses state-of-the-art out-of-core algorithms for almost every task. Therefore, you do not need to worry about your computer memory. 

You can easily register an unlimited number of images and/or laser scans on a single machine. However, it requires a specific workflow using components. In such case, 16GB should be enough for thousands high resolution (36-80 MPX) images. You can find more information about the component workflow in the application Help, section ‘Using components’.

Meshing, colouring and texturing are completely out-of-core. This means that once you have even one million images or scans registered together, you can create a mesh and a texture on a machine with e.g. 16GB of RAM without any performance loss.

However, if you need to deliver results in a limited time, you can divide the reconstruction into several parts and reconstruct the scene on more computers in parallel.

Advantages: 

• You can compute the scene by parts on different computers in parallel.
• The adjacent models can be joined smoothly.

Disadvantage: 

• The blocks need to be united into one mesh outside RC.

Prerequisite: 

• A project with the final component.

You now have 2 options - to:

a) divide the reconstruction process before computation of depth-maps, like this:

1. Open the project. Define a global reconstruction region. Export it.
2. Define some sub-regions. You can create a script for it outside RC for processing the exported region. Or you can use ‘RECONSTRUCTION tab/Tools/Info Panel’ to define them with numbers. Export each sub-region to a file.
3. Copy the project and source images for each worker.
4. Distribute the sub-region files among your workers.
5. Open the project on each worker’s PC, import another sub-region and start a model computation.

If you started computation of everything on one PC and the model computation is after 50% (i.e. the depth maps are already computed) and you decided to compute it by blocks, then you can re-use the cache and speed up the blocks’ computation on each PC twofold.

b) divide the reconstruction process after depth-maps computation, this way:

1. Pause the computation and wait until it is paused.
2. Back up your whole cache, no matter how large.
3. Distribute the cache to all PCs of your workers.
4. Continue with the steps mentioned in the section a).