IN THIS ARTICLE
Recommended
- Ubuntu 18.04 64 bit or Windows server 64 bits over Windows 10.
- CPU quad-core or hexa-core Intel i7/Intel i9/Threadripper/Xeon/.
- GeForce GTX 1070 and up (compatible with OpenGL 3.2).
- Storage: SSD.
- Small projects (under 100 images at 14 MP): 8 GB RAM, 15 GB SSD Free Space.
- Medium projects (between 100 and 500 images at 14 MP): 16GB RAM, 30 GB SSD Free Space.
- Large projects (over 500 images at 14 MP): 32 GB RAM, 60 GB SSD Free Space.
- Very Large projects (over 2000 images at 14 MP): 64 GB RAM, 120 GB SSD Free Space.
- Swap memory to be set to the generally recommended values based on the RAM size and the OS type.
High-end Build Components
Information:
For applications where low processing times and large datasets require the use of high-end workstations, there are high-end components that have proven to work better than others in most cases. An ideal high-end workstation provides balanced resources to avoid bottlenecks that slow the system down overall. Below are some of the components which have performed well with Pix4D.
Example:
- CPU: Threadripper 1950x - Core i9 9900K, Core i7 7980XE.
- GPU: GeForce GTX 1070, 1080 Ti.
- Storage: SSD.
Information:
- The graphics card may have an improvement on the processing speed for step 1 if the graphics card is compatible with CUDA (NVIDIA Graphic Cards). Processing time of step 2 and step 3 are not affected by the GPU. The GPU also helps considerably with rayCloud visualization. For more information about the use of the GPU: Use of the GPU in Pix4Dmapper.
- For more information about hardware components usage when processing with Pix4Dmapper: Hardware components usage when processing with Pix4D.
- For more information about processing speed: Processing speed.
- Unusually long times have been observed when processing with some Quadro GPU. This is related to some settings in the NVIDIA control panel. For more information: Long Processing Time for Step 3 with Quadro GPUs
- GeForce is not supported on Windows server platforms. Professional server solutions must
be selected instead. Keep the GPU drivers always up to date. - Using multiple GPUs for processing is not supported in the current version.
Hardware case study
We have performed an experiment to investigate the total processing time using a dedicated machine for Pix4Dengine SDK processing:
- Python 3.6.7, 64-bit.
- Ubuntu 18.04.3 LTS 64-bit.
- CPU: Intel(R) Core(TM) i7-5820K CPU @ 3.30GHz (6 core / 12 logical).
- RAM - 64GB.
- Storage: local SSD (no LAN/WLAN transfers).
| Inputs | Pipeline configuration | Outputs | Time |
|---|---|---|---|
|
2200 images (SONY DSC-RX1R_35.0_6000x4000) Agriculture area - large uniform fields No GCPs |
Full (all 3 steps) Additional configuration:
|
Densified point cloud (.las) Digital surface model (.tiff) Orthomosaic (.tiff) |
Calibration: 2 h Total: 18 h |
Important: Docker or any other virtualization/containerization is not part of the Pix4Dengine (although it may run in it) and so it is not included in Pix4D support. Make sure the HTTPs requests (our communication also uses TLS1.2 for message encryption) make it through your platform to https://cloud.pix4d.com/ All the protocols we use are all 100% industry standard with no customization. We do not recommend using deep packet scanning.
Article feedback (for troubleshooting, post here instead)
0 comments