Connecting 3D/360° Virtual Tour Buyers to Service Providers powered by
the Most In-Depth, Relevant and Up-to-Date Content and Training for the 3D/360° Community
Join FREE! 50+ Benefits
Should I buy a Camera? Quiz How Much to Start 3D Tour Business? Calculate How Much Can I Make as a Pro? Calculate Win More Listings? Calculate
Option #1
Find a 3D Tour Photographer Map
We Get Around Network Find a Pro Service
Matterport Service Provider? Join We Get Around Network
Option #2
Book a Matterport Pro in 60 Seconds
We Get Around Network Instant Booking Service
Powered byKoaWare
Matterport Service Provider (USA)? Join KoaWare Free! Register Now
Last 30 Days: 49,284 Page Views | 12,179 Visitors | 79 New Members
WGAN Knowledge Base | Total Posts: 60,248  |  Total Topics: 9,321
5,179 WGAN Members in 129 Countries
We Get Around Network Forum
WGAN Training Academy (Learn more)
WGAN-TV Channel
Intro 101 to Matterport for Pro Photographers
with Matterport's Jonathan Buckley and Amir Frank
WGAN-TV: Coming Up
Live at 5
5 pm EDT (GMT -5)
Tuesday, 27 August 2019
Add to My Calendar WGAN-TV Program Schedule | Need Help? | All Videos
AccuracyAECMeasuringSPAR3D17

Accuracy of Matterport8784

ron0987 private msg quote post Address this user
This question has popped up a few times and I have seen samples of Matterport compared to laser scanners and the data being quiet good and has pushed the market towards AEC (architecture, engineering, construction) trade.

I am now worried after reading an article that interviewed the MP project manager saying that it is more of a tool for 3D visual representation since it has up to an 8% error. I have included this link from an interview for Spar 3D which is a great resource for 3D scanning.

So if you are using it for AEC what error are you disclosing. I was happy with some of the past posted comparisons but are rethinking it. The other concern I have is we are trying to break in to the more commercial side of real-estate market which base the end customer maybe using the model to set lease rates. At an 8% +/- could be a big deal.

I posted this more for information because the accuracy question pops up every once in a while and now it is posted or confirmed by Matterport
Post 1 IP   flag post
WGAN Forum
Founder
DanSmigrod private msg quote post Address this user

WGAN-TV: Matterport Business Development Manager Tomer Poran AEC Market Presentation at the SPAR 3D Conference in Houston on Monday, 3 April 2017. Tomer says: 99 percent to 99.5 percent accurate of the point cloud. [4:13 into video]

@ron0987

My understanding from reading the SPAR3 article is that the eight (8) refers to the use of CORTEX for 360º photo spheres shot with a non-Matterport scanner such as a Ricoh Theta V or Insta360 One X camera.

My understanding is that Matterport claims - and still sticks by this - a one (1) percent accuracy when using a Matterport Pro 3D Camera.

For the benefits of others reading this discussion, here is a paragraph from the SPAR3D article quoting Matterport senior business development manager Tomer Poran:

"Poran says the resolution of this 3D “capture” is dependent on the 360° camera you use. Naturally, a 3D capture generated from a Ricoh Theta is not going to offer the same resolution as one from a BLK360, or even a Matterport Pro. Similarly, don’t expect survey-grade accuracy, since Poran tells me that the 3D models have what Matterport has termed “an 8% max error.”"

Source: SPAR3D

In his SPAR3D 2017 (SPAR3D17) Matterport AEC presentation recorded by WGAN-TV and posted here in the WGAN Forum, Tomer says: 99 percent to 99.5 percent accurate. [4:13 into video]

@ron0987 Your follow-up thoughts?

Best,

Dan
Post 2 IP   flag post
WGAN Forum
Founder
DanSmigrod private msg quote post Address this user
@ron0987

A related WGAN Forum discussion from three years ago:

Accuracy of Matterport Scan for Construction

best,

Dan
Post 3 IP   flag post
ron0987 private msg quote post Address this user
@DanSmigrod

This is the quote I read in the article "Matterport fed that data to a neural network and trained it to recognize how the RGB data for an object relates to that object’s depth measurements. As a result of this training process, Cortex got surprisingly good at taking 2D images of a space and inferring what it looks like in 3D, with a “max error” of 8%."

This is what I was referring to and you maybe right I worry about what they are saying and what we are actually getting, this is the second take away from the same article.

"Who is it for?
He readily acknowledges that this puts the system at the low end of the accuracy spectrum. However, if you know Matterport’s Pro 2, you also know that the company doesn’t see this specification as a stumbling block. Poran emphasizes that the system offers a very low price point, a quick workflow, and results that are more than accurate enough for visualization applications and very basic measurement needs."
Post 4 IP   flag post
WGAN Forum
Founder
DanSmigrod private msg quote post Address this user
@ron0987

When Matterport Senior Business Development Manager Tomer Poran is quoted about accuracy of eight (8) percent, I believe he is referring to Cortex created Matterport 3D Tour using panos from a 360º 1-click camera such as Ricoh Theta V or Insta360 One X: not the Matterport Pro camera.

Can you write Matterport Support and ask for clarification by Tomer and post the results here?

Thanks,

Dan
Post 5 IP   flag post
xavierchardon private msg quote post Address this user
Hello,

You might be interested by this comparison between Matterport and Leica BLK.




The space is small but we have significant differences, over 1%.
Post 6 IP   flag post
ron0987 private msg quote post Address this user
@xavierchardon Thank you for the post and the information, this is the concern I had and one of the things that worries me most as you pointed out this is a small space and if the error is more than 1% what will it be if the space is larger or the space is more complex.

Dan I have been gone all weekend but will send MP a request.

Ron
Post 7 IP   flag post
Basic
Member
thorn604 private msg quote post Address this user
The way I read it, the article clearly states the 8% refers to Cortex's "AI" inferring 3D data from the 2D 360 images captured by a Theta V or One X.

The Pro and Pro2 is still claimed to be +/- 1%.

@xavierchardon What was your workflow for that comparison?
We just completed a ~5,500sf scan using the BLK360 for the entire exterior, including roof, and the Pro for the interior. When combining the two pointclouds to build our as-built model it needed an adjustment of less than 1% to accurately mesh with the BLK360 Pointcloud.
Do you have a higher resolution image of your comparison? I'd like to take a closer look at it (the image above is unreadable
Post 8 IP   flag post
xavierchardon private msg quote post Address this user
@thorn:
2 seperate Capture project:
-with Matterport camera
-with Leica BLK

3 kind of processing:
-Matterport point cloud with Matterport camera
-Matterport pointcloud with Leica BLK
-Recap pointcloud with Leica BLK.

See the processing comparison:





On this project I didn't use the Matterport camera and Leica Blk on the same Capture project.

Comparison of the file above was between Matterport point cloud with Matterport camera vs Recap pointcloud with Leica BLK.

We used Recap pointcloud to create a Revit model.

Maybe the fact you use both on your project helped Matterport to better scale the model.....

Send me your email adress as a private message for the HD pictures
Post 9 IP   flag post
Basic
Member
thorn604 private msg quote post Address this user
Thanks for the details @xavierchardon.

Our work flow on this project was similar. Exterior was scanned using the BLK360 with MP Capture, while another Technician was inside scanning with an MP Pro and Capture.

Full resolution scans were then extracted from the BLK360 using Recap Mobile (the BLK360 app is garbage and should be destroyed), and registered in Recap to produce the exterior pointcloud. (The only reason we used Capture on site was to give the client the MP dollhouse view, otherwise we would have used Recap on site).

The interior scans were then ordered as a Matterpak and the .xyz pointcloud was processed in Recap to produce another .rcp. Both pointclouds were then overlayed in AutoCAD and using the more accurate BLK360 data as the reference, the MP Pro pointcloud had to be adjusted slightly less than 1% to fit properly.

So our workflow was similar to yours. Perhaps our Matterport Camera just felt more confident knowing that it's BLK companion was nearby scanning the same project....
Post 10 IP   flag post
60988 10 10
This topic is archived. Start new topic?
destitute