Pix4Dmapper can be run in the Azure cloud where users can take advantage of large servers for image processing.
Image processing can be performed by accessing Pix4Dmapper via Microsoft Remote Desktop from any Windows or Apple computer. However, due to OpenGL limitations the rayCloud tool is not accessible from Azure.
If rayCloud access is required, the dataset can be moved after initial processing to a local machine and further processed locally.
Guidance on Virtual Machine size
The table below recommends four Azure VM instance types which are quality tested options for Pix4D processing. These are ranked from least to most expensive. Note that not all cores will be leveraged during all parts of processing. The 32 core instances will not be twice as fast as 16 core instances, but overall will reduce processing time significantly. A10 and A11 are recommended as a good price/performance balance for typical workloads.
| Instance Type | Cores | RAM | Notes |
|---|---|---|---|
| D3 | 4 | 14 GB | Performance similar to a modern quad-core desktop |
| A10 | 8 | 56 GB | Significantly faster than a quad-core desktop |
| A11 | 16 | 112 GB | |
| G5 | 32 | 448 GB | Fastest processing available at high costs |
When considering the price for a virtual machine instance you should be aware that you are only charged while the machine is running. After processing you can stop the virtual machine via the Azure Portal and you will not be billed for compute usage until it is started again. Note that shutting down the computer via Remote Desktop does not stop billing, you must use the stop button in the Azure Portal.
Processing:
Everything is available as on the desktop computer (processing, Map View, Mosaic Editor, Index Calculator), except the rayCloud 3D interface and CUDA computing that are not currently supported in Azure.
Instructions to run Pix4Dmapper on Azure:
1. Create a virtual machine instance in Azure and connect via RDP.
2. Copy Pix4Dmapper installer and your source data to the virtual machine using one of the following methods.
2.a. A good option for large datasets is to modify the RDP connection file you downloaded from Azure to enable local drive sharing. Then you will see your local drives as network drives when logged into the remote machine.
2.b. For smaller datasets, simply performing a windows copy/paste from your local machine into the remote desktop session works well.
3. Install and launch Pix4D on the remote machine
4. Perform processing
5. Copy processed data back to your local machine using one of the options described at step 2.
6. Stop the VM in the Azure Portal when complete, to halt hourly billing.
Now that the N series Azure machines are available with Nvidia Tesla K80 GPUs, is it possible to get around error e0094 and open rayCloud on the Azure machine? It would be really useful for a project that a client of ours is working on.
Hi Ian,
We do not guarantee and support the connection to virtual machines.
However, we know that it is possible to visualize the rayCloud.
Please follow our instruction on Error e0094: It was not possible to initialize OpenGL where we give some guidelines.
Regards