Cyark Discovers Extreme Power in HP Z8 for Digital Archiving
In a comparison study, the nonprofit organization found the HP workstation far outperformed its own heavy-duty machine.
Preserving Culture With
Immersive Experiences
Share on
The world continues to remain in danger of losing its cultural heritage due to natural as well as manmade factors. CyArk works to preserve digital replicas of these structures, with an intent to preserve and educate society. The nonprofit organization aims to permanently record these significant historical monuments by creating immersive experiences. How exactly? CyArk processes extraordinary amounts of data by providing these sites with various deliverables including animated 3D models, engineering drawings, and detailed maps. In an effort to realize increased efficiency and improved workflow, CyArk put the HP Z8 G4 to test, comparing it against Titan, its own workstation. Both leveraged data from a large and complex project: Mexico City’s Metropolitan Cathedral.
Industry
Arts & Culture
Objective
Test the performance of the HP Z8 G4 workstation against CyArk’s Titan, the organization’s highest performer
Approach
Use the most complex data set—the Metropolitan Cathedral in Mexico City—to examine the compute possibilities and superior performance within the HP Z8 G4
Searching for a Faster Way
Creating a 1:1 digital replica of a cultural site involves the collection of sizable amounts of data. Since CyArk targets conservation, recovery, and discovery, it works to create a comprehensive record of sites—spanning 3D models, architectural drawings, and more—to meet these goals.
Founded in 2003, the lean organization with a team of eight has documented more than 200 sites on seven continents.
Some examples include the Sydney Opera House, Skara Brae, a 3000 BC Neolithic settlement in Scotland, and Mount Rushmore in South Dakota.
To acquire accurate data of these structures, CyArk deploys LiDAR or laser scanning to measure the distance to an object and map its location, as well as photogrammetry, a series of overlapping photos, to generate a 3D model.
“The Z8 was able to process all the data together at once and then, upon visual inspection later, we saw that it did give us better results”
Scott Lee
Director of Reality Capture, CyArk
“We use this type of sensor fusion because it gives us the best photo replication possible,” says Scott Lee, Director of Reality Capture at CyArk. “By blending the two technologies, LiDAR and photogrammetry, together we can create the most accurate representation.”
When the organization acquires LiDAR data, it results in billions and billions of data points. The team pre-processes the data by registering or linking the scan data, followed by color-correcting both the aerial and terrestrial photos. From there, it moves into RealityCapture, a photogrammetry software that combines the datasets and ultimately reconstructs a 3D mesh model of a site. Based on a project’s deliverables, CyArk leverages game engine technologies to create animations or VR experiences. In addition, it creates drawings for conservation or simplifies the data for a viewing platform to be publicly accessible.
The sheer volume of data takes several manhours for a project to reach completion. CyArk had been processing data in chunks and then stringing them together.
“By manually tying the data together, we were losing all the automation and algorithms that make the data really nice and allow it to become one cohesive dataset,” Lee says. “We knew we could get better results.”
IT Business
Objectives
• Test compute power of the HPZ8
• Process 13,252 inputs and 796 GB of data, a 295 GB point cloud file, 1,257 (100. mega-pixel) medium-format, 2,219
• (36 mega-pixel) DSLR, and 6,007 (20 mega-pixel) drone photographs
• Create a textured 3D model with maximum resolution
• Develop a simplified, textured 3D model targeting 40 million polygons
• Realize cleaner and more detailed output on final model with fewer artifacts
Business
Objectives
• Significantly save processing time and man-hours
• Reduce need for manual adjustments to the model
• Increase accuracy of models
• Leverage a high-performing workstation for future complex, large-scale projects
About Cyark &
Partners
CyArk is a nonprofit organization founded in 2003 to digitally record, archive, and share the world's most significant cultural heritage and ensure that these places continue to inspire wonder and curiosity for decades to come. CyArk assists preservation efforts, captures comprehensive records of sites, and shares data in powerful ways to convey the importance of these places. It has documented over 200 sites on all seven continents.
Comparing performance
To see if it truly could get better results, CyArk chose to compare the power of the HP Z8 G4 against Titan, its highest-performing machine.
The HP Z8 G4 selected for the project featured this configured solution:
• HP Z8 G4 1125W PSU
• Dual Intel® Xeon® 6240 2.60GHz 18C CPU
• Dual NVIDIA® RTX 8000 graphics with 128 processing cores
• 384GB (12 x 32GB) DDR4 2933 ECC registered memory
• HP Z Turbo Drive M.2 1TB TLC SSD
• 2 x 4TB 7200 SATA Enterprise 3.5in
• Premium Front I/O (2 x USB3.1 Type C, 2 x USB3 Type A) The Titan, also with an Intel Xeon processor, had the NVIDIA Quadro P600 GPU with 44 processing cores.
CyArk selected one of its most complex and largest datasets—the Metropolitan Cathedral—to run the comparison study. A world heritage site, the cathedral is located in Mexico City’s historic district.
“It presented itself as the perfect project to test,” Lee says. “We had metrics and knew what the dataset took prior because we were recording the steps all along.”
The dataset was loaded on the two machines and put through the same motions that CyArk uses to generate 3D models. The machines processed 13,252 inputs and 796 GB of data, a 295 GB point cloud file, 1,257 (100 mega-pixel), medium-format, 2,219 (36 mega-pixel) DSLR, and 6,007 (20 mega-pixel) drone photographs.
The team figured it could throw all of its data into the HP Z8 and based on the expected performance of the workstation, it would be able to manage it.
“Our assumption was correct,” Lee says. “The Z8 was able to process all the data together at once and then, upon visual inspection later, we saw that it did give us better results.” Inspection included checking for areas of misalignment in the data. The HP Z8 took five hours to import and align inputs compared with Titan’s 16 hours, over 70% faster.
Customer At a Glance
Application
Preservation of cultural heritage through the collection of aerial and terrestrial data to render animated 3D models of structures.
Hardware
HP Z8 G4 Workstation
“Every time something is not perfect on the Titan, you have to do further refinement,” Lee says. “But to be able to get those results after five hours initially with the Z8 was kind of mind-blowing.”
In nearly every step of the journey required to generate digital replicas, the HP Z8 outperformed Titan.
With the HP Z8, reconstruction took 1 day and 21 hours compared with Titan’s 16 days and 20 hours. In total, the team saved 14 days and 33 hours in compute time.
“Every time something is not perfect on the titan, you have to do further refinement. But to be able to get those result safter five hours initially with the Z8 was kind of mind-blowing”
Scott Lee
Director of Reality Capture, CyArk
Overall, the team ended up with cleaner and more detailed output on the final model of the cathedral, including fewer artifacts, which would also lead to less personnel time if the model were taken into other visualization formats such as virtual reality.
In terms of personnel time taken on the project for tasks such as preparing datasets, setting texture settings, and creating orthophotos, the savings were significant. The HP Z8 took 13 hours while Titan took 220 hours, saving more than 5 weeks of personnel time.
“For a little organization like ours, we’re a small nonprofit, our time is super valuable,” Lee says. “It’s kind of funny, you know the Staples ‘Easy’ button? It’s pretty much what happened.” The project results: an interactive 3D model and 3D animations of the building’s exterior, exterior to interior, and a cut-through.
Setting the stage
The comparison study helped Lee recognize that his team could process data with ease. On average, CyArk runs 12 projects a year. Having the HP Z8 on hand has changed the organization’s workflow for the better.
“Normally what we did in our workflow was essentially one person would come in and work their way through the data on one of our three better computers, Titan being one of them,” Lee explains. “They would work their way through the reconstruction of the model in RealityCapture and then generate drawings or simplify the data, getting it into a Unity engine or some other game engine for virtual reality.”
Now, he says the team uses its other computers to do the pre-processing, content creation, and other deliverables like drawings. The HP Z8 is reserved for actual reconstruction and model-making.
“It makes it much more efficient because we can have three or four workers all working on the other steps that take a lot of human compute time,” Lee says. “And then we can push those results into the Z8 and much more quickly and accurately have the model. We can then send it back to different team members and get the results we need for other deliverables.”
The HP Z8 offers CyArk confidence in what it can accomplish, he says.
“We used to limit data captures of sites in the field due to post-processing time,” Lee says. “The Z8 helps eliminate that and allows us to capture much more.”
Z by HP for Data Scientists & Analysts
Get rapid results from your most demanding datasets, train models and create visualizations with Z by HP data science laptop and desktop workstations.
Design the spaces of tomorrow with NVIDIA RTX™ graphics.
Exceptional performance
with Intel® Xeon®
and Intel® Core™ i9 processors.
You May Also Like
• Case Study
How is NASA Monitoring The Surface of The Sun?
• Trend Article
Z by HP Data Science Workstations and Software
• Trend Article
AI for All: Machine Learning as a Service
• Case Study