GeoCV is real competitor Masterport!

New York City

AntonYakubenko private msg quote post Address this user

Quote:

Originally Posted by JuMP
@AntonYakubenko How did you get 3D mesh of the house? Is it by photogrammetry from the photos?

We are working long time in photogrammetry area. And we found photogrammetry is good solution for some object. But it is can't handle all materials within a house. The Matterport way (with PrimeSense's sensor) is the best way with best cost performance.

We're using active depth cameras like PrimeSense for capturing, in particular, PrimeSense itself, Occipital Structure sensor (based on PrimeSense Capri), built-in 3D cameras in Lenovo Phab 2 Pro and Google Tango mobile devices, and tried others as well like Intel RealSense. I totally agree that using such kind of sensors is the best way to go for scanning interior spaces.

And you're absolutely right about photogrammetry. It works great for textured objects, but completely fails in real interior spaces with flat white walls, for example. Below are a couple illustrations.

A great 3D reconstruction of a very textured scene (courtesy of realities.io, processed by Capturing Reality software):

A failed 3D reconstruction of a more usual room (processed by Autodesk 123d software).

There are options to solve this problem by incorporating more high-level knowledge into 3D reconstruction algorithm, for example, recognizing edges, fitting in walls, as well as adding some manual input. In my understanding, that's what InsideMaps and Real.Vision are working on. But from what I saw so far such an approach doesn't produce really high-quality and complete 3D models in a robust way.

Another way is to fuse robust 3D measurements from a depth sensor with higher resolution photos, which also work outdoors, to get the best of the both worlds.

Quote:

Originally Posted by JuMP
The mesh result from your demo is very abnormal compare to Matterport. I guess it was caused by UV issue of your textures. You sould seperate the object mesh at the seams on UV.

Thanks for the feedback. We'll check the mesh, but it seems you've used a hack to show a wireframe model in our WebGL viewer and it's not shown correctly because vertices are connected in wrong order. You can see the real mesh here: https://skfb.ly/6nAYA

And here is the screenshot:

Quote:

Originally Posted by JuMP
And the accomplishment rate of your DEMO looks like our POC.
You can visit our POC for richtag DEMO of Matterport here:
clickable text

What do you mean?

Quote:

Originally Posted by JuMP
I like Matterport. And I want to give Matterport an advice:

Add another purchase option:

1. Lower the hardware price from USD3600.00 to 120% of the cost of it:
For example: if the materail cost is USD1000.00 then sell at USD1200.00
2. Higher the data process fee from USD19.00 to USD49.00 for the first 100 showcase.
3. After the user shoot 100 showcase that means USD4900.00 total, the data process fee will lower to USD19.00 again.

The KEY for Matterport's business is to make more his cameras in market and get more data process and more users/clients. Not to get the revenue of selling camera. The higher hardware price give more area for his competitors. But actually the hardware cost of Matterport won't more than USD1000.00 by my estimate. As a network/cloud services business company Matterport should expand his cloud computing business which will get more revenue because the cost of machine processing is linear and its revenue is exponential.

It's not my business to count other companies money, but Matterport has sold over 5000 cameras for $4,500 each, which gives them $22.5m in revenue for hardware, and 400,000 scans approximately for $19 each, which results in $7.6m for software subscription.

But our thoughts were similar to yours and we're taking a different business model approach.

7 years ago

Quote:

Originally Posted by AntonYakubenko
...a very similar experience has been provided before that by Google Maps (not Google Business View), for instance. There is a true volumetric 3D model of the city, which you can rotate and pan freely. And then you can zoom in or jump into street view panoramas. The transition between panoramas are also smooth because Google also uses the underlying 3D mesh. You can even see the same 3D cursor changing its orientation depending on the normal to surfaces. If you go further you can actually notice similarities in the layout and design of controls as well.

I did not realize that Google was integrating the 3D Mesh. As both an MSP and a GTP -- I have found that Google's transitions and navigation is horrible and am surprised that the 3D Mesh is in use ... the transition between scenes certainly does not feel as smooth or look as nice as Matterport and that might be due to the mesh being less detailed as we are, of course, looking at a massive model with less detail at street level than Matterport at camera level. Thanks for pointing this out!

Quote:

Originally Posted by AntonYakubenko
Why GeoCV, Google Streetview and Matterport 3D viewers look similar? Because they are using the same type of data and all have independently found that it's the best way to visualize it.

Agreed. There's only so many ways to skin a cat. I think that the feature set certainly determines the UX. One decision that I'm pleased you have made with GeoCV is to have the Floorplan and Dollhouse views as separate buttons. Matterport's combination of these two items into one button was a decision that I vocalized my dislike for. People end up NOT finding the floorplan if that's the case -- they are too busy with the dollhouse, then usually click into it and begin navigating again. I think that bringing new features to the market first and integrating them into the UI would certainly allow you to differentiate a bit.

Quote:

Originally Posted by AntonYakubenko
But we see some opportunities to improve the navigation and viewing experience and to add a unique look and feel to our product. So stay tuned for the upcoming updates.

Great!

Quote:

Originally Posted by AntonYakubenko
(I like the pre-visualization of panoramas in small circles done by ImmoViewer and RoundMe).

Those are extremely slick touches -- I think they make a big difference.

Quote:

Originally Posted by AntonYakubenko
But in contrast to Matterport we use 5x-10x cheaper hardware delivering up to 15x higher resolution of imagery.

That will be your defining quality. If that quote can stand up and is the experience that your users have, then you have an excellent produce and this will be successful.

My only concern, which I expressed to you, was patents that Matterport owns and whether or not your solution would require a licensing fee from Matterport in order to exist or offer certain features.

9.44K Frisco, Texas	9.44K Metroplex360 private msg quote post Address this user	plus 1 plus 1 plus 1 3
	Quote: Originally Posted by AntonYakubenko @VisualKusuma, thanks for the feedback. Quote: Originally Posted by VisualKusuma My only wishlists on this would be.....white label Matterport allows some branding. Is it not enough? Do you want to remove the technology provider branding at all? Doesn't it add some value as soon as the technology provider brand becomes strong? What about our modest light text "Powered by GeoCV"? Quote: Originally Posted by VisualKusuma & no subscription plan What would be the more suitable pricing option? I'm on the side of keeping the branding -- as long as it is not intrusive and distracting. I'm also a fan of the subscription plan as companies need to make $, pay their employees, build infrastructure and continue to innovate. From the consumer side, maintaining a strong cloud-based platform for fast delivery of content is necessary and allowing tours to be served locally only creates problems, fragmentation with results, and spurs a new 3rd party hosting market. Matterport are charging $50-$150 a month which is not bad considering that the price is $0 for hosting if you are scanning 3, 7 or 11 tours per month. No one should really be complaining about hosting except when and if they decide to stop scanning -- at that point, there's a bit of a problem.
	7 years ago	Post 30 • IP flag post

GeoCV Co-Founder 686 Website New York City Linked In Facebook Twitter	686 AntonYakubenko private msg quote post Address this user	plus 1 plus 1 plus 1 2
	Quote: Originally Posted by Metroplex360 One decision that I'm pleased you have made with GeoCV is to have the Floorplan and Dollhouse views as separate buttons. Matterport's combination of these two items into one button was a decision that I vocalized my dislike for. People end up NOT finding the floorplan if that's the case -- they are too busy with the dollhouse, then usually click into it and begin navigating again. Oh, thanks. It was an intuitive decision for us. I didn't know Matterport users are experiencing problems with not seeing the floor plan.
	7 years ago	Post 31 • IP flag post

GeoCV is real competitor Masterport!4313