I had to expand on the previous, so here is Matterport Capture Compare 2.0!12070
Pages:
1
|
RichardStanton private msg quote post Address this user | |
| Greetings all - I decided to make a follow-up test to one I had done earlier this month, and now that I have completed the execution I figured I would also share my findings. Now, this is hardly a scientific test - but it is also not completely uncontrolled. I am a very pragmatic person, and as such I rely on empirical evidence quite a lot for bridging the gap of data sheet to reality. As I am sure I am not the only one like that, I figure why not share. Answering the Why As all of us know the device factor has been steadily increasing in this area, and not just Matterport itself. The discussion of device is becoming more common, and because quality is such a subjective thing I have to imagine that coming into this new off of the street would be rather frustrating and full of biased opinions. Throwing into the mix we are now not only talking about different devices, but different devices using fundamentally different methods of capturing data, each of which have their own strengths and weaknesses. Hemispherical 360 capture for instance is extremely fast, but can lack on the quality side due to the huge FOV on a shared sensor. Pano/Stitching is obviously slower, but has the potential of a much higher resolution and quality end result, if one assumes that the stitching method is of reasonable quality. Then we have the photogrammetry approach of tracking movement vs. direct point cloud scanning, or some hybrid of the two. Head.. spinning. This tiny test will not answer those questions - but it will at least show the current state of a few of these methods and their empirical results. Unfortunately I do not have access to a Theta or BLK360 to round out the example list, but partial is better than none! Since I am on a grandfathered account structure, without a complete device matrix to test I was not inclined to incur charges for each of the test scans in a larger sample size. As such, two scans each. Standardization/control factors * All devices as of the time of this test had the latest firmware, patches, and the same version/build of the Capture application AND the Cortex revision. * All devices were used as per their primary use case. For this test it means that the Pro 2 and the Insta360 One X had stabilization assistance via tripod and monopod respectively. All iDevices were hand held and performed a BASIC scan. I chose the Basic scan because it is the closest in time of execution to the other methods, and likely right at the limit of time most people would spend. Detail scan for those who have not used it not only does a single 360 grab for pano creation, but it also requires a second and THIRD spin around to capture the + and - elevation angles for a larger area of view. This is decidedly NOT fast, though I may introduce detailed captures in future tests as a measure of IQ (Image Quality) and immersion. * I was COMPLETELY sober for each scan, though it might be hard to believe based on some of the results!  Tested Devices * Matterport Pro 2, iPhone X, iPhone 11Pro Max, iPad Pro 2020/4th Generation, Insta360 One X (again sorry, no BLK360 or Theta hardware  )Okay, that's it, I will shut up now, and close with the test models and a rather popular quote: "This above all: To thine own Eyes be true"... okay maybe it didn't go EXACTLY like that... iPhone X: https://my.matterport.com/show/?m=kfSg3afzE5y iPhone XI Pro Max: https://my.matterport.com/show/?m=KcQRbiPiRkj iPad Pro 12.9" 2020/4th Generation: https://my.matterport.com/show/?m=jPgt1Wfwc8v Insta360 One X: https://my.matterport.com/show/?m=K5dsU3h4rVW and last but not least - Matterport Pro 2: https://my.matterport.com/show/?m=dFQaMDFvQn2 I hope that this is found useful for at least a few! |
||
| Post 1 • IP flag post | ||
|
rastas private msg quote post Address this user | |
| Thanks for this. Its intersting to see how bad the results are on everything other than the matterport pro 2 (relatively speaking). Means the matterport camera I have is still safe for a while at delivering the best quality. | ||
| Post 2 • IP flag post | ||
 WGAN StandardMember Lisbon, Portugal and London |
Pedrotex69 private msg quote post Address this user | |
| Thanks for the tezts. To be honest only two are pretty much in the competition, the Matterport Pro2 and the Insta360 One X. Not knowing the real colour of the cupboard I can not tell which one came closer to the real colour. But if I was about to buy a camera and looking at the difference in prices between them and all that all that is going on with Matterport, I probably would go to the Insta360 One X and not the Matterport Pro2. Actually at the moment I am facing this dilemma, I have an Insta360 One X and I am thinking in buying another camera and one of the options is the Matterport Pro2 or another 360 camera. Can anyone advice me which camera I should buy and why? All the best and keep safe. |
||
| Post 3 • IP flag post | ||
 WGANFan Club Member Cairo, Egypt |
AhmedAttia private msg quote post Address this user | |
| Amazing example and good proof for idiot customers whom just want a virtual tour regardless what was behind and how will be the delivered quality. Well done. | ||
| Post 4 • IP flag post | ||
|
|
RichardStanton private msg quote post Address this user | |
| To be fair I will probably introduce the high quality scan into the mobile tests - this would better approximate the amount of data as well as the field of view that both the Matterport II and the Insta360 enjoy. The primary reason I did not do that is due to the amount of time it takes to actually do it. I simply cannot see anyone wanting to spend that kind of time. The only thing I see bridging the gap would be to introduced a wide angle accessory. I also did not look closely at which of the lenses Capture was using for the two Pro devices. Both pro devices come with an integrated ultra-wide lens, and I did not notice a drastic difference in the number scans to comprise the 360 that I would expect in the difference between the iPhone X standard lens and the Pro ultra-wide were it being used. I will take a closer look to verify - but that could mean that the gap would most definitely narrow through future Capture updates. @Pedrotex69 there really is not an easy call as to which way to go, but if you integrate your own goals and the prospective job types you are looking to gain it will likely send you down one path or the other. Namely if you are looking to do more commercial spaces, or work with builders for as built or general BIM bolstering, etc. then the MP2 is much more capable in these areas. There is a reason that Matterport does not support floor plans taken with 360 camera jobs and it's a simple one - their accuracy variance is way too erratic to be trusted with anything important. So, I hope that helps nudge you in whatever direction is best for you and your business!  |
||
| Post 5 • IP flag post | ||
|
|
Changesin3d private msg quote post Address this user | |
| It was my understanding that Matterport was farming out the floor plans to Cubicasa that could be the reason they are not doing it with the Iphones. I have seen a video that shows the Cubicasa system working with the phone. Maybe that is the reason. Surprising that they are going this direction and still calling the lower grade scans the same thing. This leave it to the good camera owners to explain the difference. Is lumping all of the options together really fair to the guys who have the best equipment? | ||
| Post 6 • IP flag post | ||
|
|
ahojman private msg quote post Address this user | |
| Richard, 2 comments and a question: 1) Interesting comparison. 2) you make me lost a lot of money because now my wife wants a kitchen like yours. 3) if you work with the iphone mounted in a tripod and rotating 90 degrees 4 times, does Matterport stitches correctly the 4 photos? I understand we also must take 3 or 4 more photos pointing to the ceiling to complete the sphere. Why do you say it takes too much time? |
||
| Post 7 • IP flag post | ||
|
|
RichardStanton private msg quote post Address this user | |
| @changesin3d Oh you CAN get measurements from a photo, but a photo based solution will never equate to a laser bouncing off of an object and being interpreted (actually a lot of lasers, many times). With their own hardware solution they have a solid foundation of data to work with which will yield predictable and proven results (hence why I mentioned commercial uses) It's pretty unlikely that you would find a builder whom would find the accuracy acceptable from a photo derived solution. Now I'm making an assumption about why MP chose not to do floor plans with these other methods, but the decision makes sense based on the reality of accuracy and their own historical CYA (Cover Your Ass) behaviors in the past @ahojman 1. Thanks! Once I am forced to move over from my grandfathered billing plan I will likely do a more thorough test with a lot more data points - I just couldn't justify paying the fee for this many scans for testing. 2. I am sorry for your loss! :'( That is not my kitchen, however, just a scan site I was able to take some extra time on since it was out in the mountains near me You must think I live an extremely spartan life style with so little furniture!! 3. I am not sure where you are getting 3 or 4 from, but as of this moment, on an iPhone XI Pro it takes 6 scan points per 360 for the Simple scan, and 18 (6*3) scan points for the Complete scan (it takes mid-line, positive x degree and negative x degree shots on each iteration) 6 takes long enough, 18 is almost painful. Stitching, at least hand held, is dubious at best right now, but I am sure that will get better with time. They have already in this short time changed the point behavior in beta. We shall see how it works out! |
||
| Post 8 • IP flag post | ||
|
|
ahojman private msg quote post Address this user | |
| Thanks for your answer! | ||
| Post 9 • IP flag post | ||
Pages:
1This topic is archived. Start new topic?