Review: iPhone LiDAR scanner + Photogrammetry | 3D Forensics CSI | CLICK 3D EP 13

[@cadamsbdabest5210]

Excellent video! I'm currently doing a forensic science research project at grad school comparing the LiDAR Scanner (+various 3rd party apps) on the iPhone 12 and comparing it to traditional scanners for scenes. Would love to get the app that you are making to include in the testing!

[@3Dforensics]

Still a few weeks away before I can get it out in public for beta testing

[@5265060]

Great work! I’ve been looking for a good comparison between the iPhones lidar and a professional level device. It would be interesting to see the same thing done with medium and small sized objects.

[@3Dforensics]

Something on the order of a couple of meters might work, but once you get down to small details, you are pretty much working with photogrammetry only and you are better off using a DSLR or using photos from the phone camera and not video.

[@stefano.ingegnere]

Amazing presentation, sir. An excellent and in-depth detailed analysis. It was very defining your video for the purchase of iPhone 13 pro max for its lidar scanner, plus the program you are using. Thank you very much

[@dougmuir2512]

I think the red in the final analysis can be blamed on the angle of incidence of the reading on the horizontal surface. Not much direct reflection. So not to blame the technology. Just know the limitations of viewpoint access.

[@FlorentPoux]

Hi Eugene! nice to see that you are interested in quality analysis coming from sensors, and nice video! A very quick comment though, C2C results is highly dubious due to the nature of the different point clouds, thus extreme care should be used before quantitatively showing deviations. I recommend switching to more robust ways to do this. Another thing, taking the Faro data as reference is also disputable, as it delivers very noisy point clouds, that will impact C2C results. Ideally, you would have the true manufacturer's mesh, or scanned with a faro arm, that you use as the base for your QC and analysis. That being said, your qualitative analysis is right on point, and it does not change a bit the quality of the video ;)!

[@3Dforensics]

Hi Florent...I wish I had access to that kind of equipment! I think for the purposes of what I am after, the FARO Scanner was accurate enough. There was some noise and other data that I could have to accept and your point is noted on the interpretation of results. There is a lot more to be done here and I admit that given more time and resources, a more controlled experiment and analysis could have been setup. Thanks for the comments and keep up your great work!

[@FlorentPoux]

@@3Dforensics Haha we all wish :). But you are right, for the purpose (showing that LiDAR and imagery is better than LiDAR alone), the FARO serves the purpose very nicely! But if, by any chance, you find such a ground-truth reference, then estimate noise ratio, registration error and surface roughness, your quantitative deviation metrics will have a whole other meaning! But you are the forensics expert thus I guess you know what is enough for this specific application :). (I emphasize again how clear your explanations are)

[@LiDAR3D]

Interesting! It would be great to see this examination done with another example than a shiny car (which you also pointed out is not an ideal thing to capture) I’m also wondering how it would look like if you did the iPhone scan correctly (literally every scanning tutorial says do not scan the same area twice and you did it 3 times on an object covered with reflective materials 🤔) As I understand it you would have spent close to 90 minutes scanning the car with the high end scanner (if you would have done it 3 times over as with the iPhone). Considering that the iPhone only takes a minute or two I think that fact could be highlighted more.

[@3Dforensics]

The EveryPoint App uses loop closure and so I was making a direct comparison. Also, I just released another video where you can see the lidar completely fails across a large building and doing only one loop or two loops. The speed of capture is a very interesting point...very easy and fast.

[@LiDAR3D]

​@@3Dforensics Jjust skipped through the new video. We quickly found that apple lidars are not ideal for exteriors, for several reasons. Kind of a no-brainer.

[@christianferrazgruner]

It seems that error you get on iphone it dues to the lack of precision in the dead-reckening of the IMU's. What if you scanned with the iphone in the same manner as the with the laser scanner, so more or less stationary on several points, and then merged?

[@jonathanmiranda7145]

Great explanation Eugene! Keep on the good work

[@3Dforensics]

Thanks Jonathan...more coming soon!

[@jonathansouthwood3767]

Hey there thanks for making this video. One of the more comprehensive ones on KZbin helpful for surveyors. May I please ask, the last model that you showed in the video of the whole parking lot, how large was the point cloud in terms of gigabytes, and what type of file formats does EveryPoint let you export in? Also, how large was the EveryPoint cloud of the car?

[@3Dforensics]

Hi Jonathan. I am pretty sure they were relatively light models and not as dense as they could be. I am on the beta version and I can request more dense models. Stay tuned as I am hoping to have an app released in the future where the user can specify the point density. You can export in e57 and ply at the moment.

[@cedricsinga3440]

Hi Eugene, thank you for this great work! I needed to make sure about the accuracy of the lidar sensor in the iPhone then I found you and you did even more. That was a great comparison, I wonder how accurate would it be scanning small objects with the EveryPoint App and make some modifications?

[@3Dforensics]

Hello Cedric. Thank you for the kind words. I have now released a photogrammetry mode in Recon-3D (my own app) and so when you need smaller details, you also get better accuracy too!

[@cekuhnen]

What about Trnio + they do photogrammetry LiDAR informed and you can also add photos from the phone or dslr to the calculation !

[@3Dforensics]

I don’t think they are fusing the same way as in Recon-3D but I could be wrong.

[@cekuhnen]

@3D Forensics yea they do - I work with their product a lot for my teaching and research. The LiDAR Scanner is for alignment and scale and photogrammetry for textures and details. Very attractive the photo set can be covering more subject than the LiDAR Scanner. Give it a try - I could see for your area this being beneficial

[@3Dforensics]

@@cekuhnen well, I am sort of committed to Recon-3D 😀

[@ericbergholz3484]

Hi Eugene, very good work. Maybe you can compare the distance values on the car. E.g. the axle centers. There may also be a value from the manufacturer. Maybe the short distance is more accurate than the distances of the parking marks. The areas are further apart and I'm not sure how big the deviation will be, since the sensor's recording range is perhaps not so large. Maybe results will be better, I would be interested.

[@3Dforensics]

Hi Eric, I think it would make sense to do more testing with smaller areas and then larger areas as well. I chose a larger set of reference points as smaller distances for reference scale points will propagate as you take larger measurements. I plan to do another test with both a larger and smaller scene and will keep you posted on the data results! Thank you!

[@uassau]

This is a very, very good video! Perfectly done sir!

[@scanningcars6500]

Really interesting - was wondering what the iphone's capabilities were.

[@3Dforensics]

More to follow in the near future...standby...

[@grootproductions9773]

Great video, Eugene! Would it be possible to import the Everypoint scan data into a software such as Blender or Maya?

[@3Dforensics]

Yes, there are people who have already done so with some plugins and such. I am not a Blender user (yet) and I use 3ds Max which is natively possible. I believe Maya is also natively possible since it is an Autodesk product.

[@grootproductions9773]

@@3Dforensics Thanks. I have tried using Point Precise with Blender, and it works perfectly fine. I'm pretty sure Point Precise is a product of Everypoint.

[@marcsawyer6749]

Have they set a date yet for the release of video + lidar?

[@3Dforensics]

I know they have an update coming out soon but I really can’t say what the exact date might be. They are working pretty hard on regular updates at the moment.

[@johnathonvought7407]

@@3Dforensics I noticed that my version of the app looks different than yours. Perhaps less controls in the everypoint+lidar options.

[@3Dforensics]

@@johnathonvought7407 Yes, I have a beta version but I may have some exciting news in a few weeks time...standby!

[@MrFabiolobo]

Very good video! Heve you ever got a Mototola G100 to test: there is a tof sensor there....

[@3Dforensics]

I have not. I have been only testing the iPhone for the moment.

[@H2Dwoat]

Hi, great video and very informative 👍. I’m interested in this as a way to capture non regular shapes to create a negative space to carve with a CNC. It doesn’t sound as if you think the accuracy of the iPhone LiDAR is that great. I have just purchased the 13promax, are you aware of any improvements to the LiDAR From the 12pro. I like that the combination of LiDAR and photogrammetry improves performance. Do you know if the inaccuracies are consistent and hence able to be accounted for? I’m really trying to get down to mm accuracy but I’ve just started researching the process and am gathering information.

[@3Dforensics]

There are a lot of questions that need To be asked like what you are scanning, and what scale? What is the surface texture like and what is the upper limit of error you can afford to live with. Photogrammetry alone can achieve high accuracy but it depends on the nature of the object being scanned more than anything. Look at some of my other videos on the Click3D playlist and I tackle a whole number of things.

[@H2Dwoat]

@@3Dforensics hi, I will definitely check out your other videos. As for the materials it will vary but I’m looking at stones initially of various types. Scale would be from coin size through to dining table but I am only at the brainstorming stage at present. Error range of 1-2mm I think would work.

[@3Dforensics]

@@H2Dwoat Photogrammetry is a great option for stones. They reconstruct really well. You just need a very accurate scale and need to take plenty of photos. Small stones the size of a coin would not be suitable for Apple Lidar. At least there is no benefit unless you were doing a larger stone. Look at programs like 3DF Zephyr, MetaShape and Reality Capture.

[@H2Dwoat]

@@3Dforensics hi, much appreciated, cheers.

[@chriscummings6495]

Thanks for the great demo Eugene. If we are looking to have a textured mesh for working in programs such as 3dsmax would you suggest continuing the workflow in CloudCompare or with a different program? Thx.

[@3Dforensics]

In some cases, CloudCompare is a great solution and pretty simple to manage. However, based on the complexity of the mesh, it may sometimes require a lot more work. At the moment, EveryPoint does not offer a seamlesss meshing method but I think that may be coming.

[@3Dforensics]

If you are thinking of using a mesh for 3D Visualization then you will definitely need something more than CloudCompare. Think about ZBrush or even Blender as they have some decent repair tools but be forewarned...there are people that make this a career and it's never as simple as it looks!

[@dilithium72]

Are the pulses from the phone triggering the car's sensors? I'm curious why the wing mirrors are moving in and out?

[@3Dforensics]

No, but the car keys in my pocket were every time I got close to the car! Annoying.

[@ckasmobile3060]

Hello Eugene, Thanks for all your testing data - it is very detailed and extremely relevant. I am thinking about using the iPhone LIDAR with Photogrammetry (EveryPoint) to scan interior of rooms during construction of houses - before the plaster cladding is on the walls (so I know where the timber studs and cabling etc is behind the walls once they are covered in plaster). I would be trying to capture "thin" objects like vertical and horizontal timbers (only about 45mm thick ~ 2 inches) and also even much thinner cables (perhaps as small as 10mm ~ 0.5"). I don't need ALL the data points to be perfect, but at least to have some confidence in a room about where things are, especially when combined with well documented photos. In light of this, and from your testing, do you think I will be able to pick up this data "per room" (not try to scan entire house)? Have you noticed if things like dangling cables or "chair legs" or similar appear in your interior scans when the LIDAR and Photogrammetry are combined in EveryPoint? I would love to see if have done this kind of scanning of "frame stage" of a house or similar! Thanks so much for your hard work!

[@3Dforensics]

I am going to say that it is certainly possible. At the moment, I am spread really thin but if I can get to something like this in the near future, I shall. The small timber beams won’t be a problem but cables and such are sometimes only partially captured so you need to make sure you have good coverage there. Thank you for the comments.

[@3Dforensics]

AndI forgot to mention…tomorrow I will be starting. Series of 3, free webinars on photogrammetry in case you are interested. www.ai2-3d.com/p4f

[@ckasmobile3060]

@@3Dforensics Wow! Thanks so much for replying so quick! Thank you also for the invitation! I will check them out! I suspected the cables would be an issue, but was thinking as long as there are "some" points and artifacts in a particular location, then coupled with good photos to look at, I would be able to determine with some confidence where the cables may be! Anyway - thanks so much once again!

[@7oppinFr3sh]

​@@ckasmobile3060 or @3D Forensics (thanks for the video!) Did you attempt the scanning of interior of unfinished rooms? If so, how did it work out?

[@HoraceHingCheungPoon]

Dear Sir. I need the configuration of your iphone. Memory etc. Also, should we use an iphone or an iPad? Thanks in advance!

[@3Dforensics]

I was using an iPhone 12 Pro, but now I am using an iPhone 13 Pro Max and and iPad Pro. 256 Gb on the phone and 512 on the iPad Pro

[@darylcheshire1618]

Would it make sense to do one pass with the iphone but 3 times longer? Then you won’t get the overlapping of the doors. I’m thinking that the LiDAR in the iphone is only good for the camera where it can measure the distance more accurately in low light rather than rely on focus.

[@3Dforensics]

The point here is that there is no overlap detection or loop closure. So, in a way, yes, you won't have 3 doors but you just don't know what you are getting.

[@enzocaputodevos]

Thanks !

[@JohnHuebbe]

When can I get this new app?

[@3Dforensics]

Actually, I am working on something...hoping in a month I will announce!

[@maheshahire100]

Can we use iPhone 12 lidar in large area mapping say about 2 Hectors to 5 Hectors to measure area and depth ,volume of excavation etc.

[@3Dforensics]

I think that might be a bit of a challenge right now and you are better off using photogrammetry.

[@tobi6758]

I wonder if you gave the iPhone a tripod and 3 min times 8 positions as well, the end result would have been better

[@sayonaradesu1087]

hello, how much memory does the scan occupy on the iphone? I want to know if 128 gb memory will do.

[@3Dforensics]

It depends on the length of scanning but I would say on the order of 100 to 500 Mb for most scans.

[@sayonaradesu1087]

@@3Dforensics oh, thats nothing. It seems that 128 gb would do. But, they let me save the scan on the phones memory right? I mean, to not force me to move the data on their cloud for more cash. I am a android user, but only iphone provides these benefits. Thats why I ask.

[@3Dforensics]

@@sayonaradesu1087 If you can wait a bit, I am working on an App that will allow you to save and process everything on your phone or the cloud. I am using a beta version that has options that others don't so I have not seen the "regular" version for some time. I am able to process on the cloud and simply download my scans from a browser.

[@sayonaradesu1087]

@@3Dforensics sounds interesting. Do you have some approximative date you can tell me on the app release and a potential price tag?

[@3Dforensics]

@@sayonaradesu1087 all I can say right now is we are in early stages with an early beta by October or end of September if we are lucky. Price has yet to be determined.

[@FPChris]

I have seen nothing that shows me that the lidar alone has good resolution.

[@3Dforensics]

Agreed, the Lidar alone doesn't get me very excited so far...I was hoping for more.

[@GustavoASMoreira]

Please analyse IPhone 13 Lidar enhancements

[@3Dforensics]

Definitely on my list of things to do…still a little while before I get my hands on the iPhone 13 though!

[@LiDAR3D]

Oh and also the fact that the FARO scanner costs 50 times more than the latest iPhone...

[@Cgboling]

The camera needs To perpendicular to var to get the best data. Those areas where it detected three different depths because rear is from the shots from the front of the vehicle shooting down the side. That angle never produces accurate data.

[@MichaelHulme]

Why do such a vaguely positioned, hastily walked scan, with only medium detail selected, with your first app run though, if you really wanted a better quality comparison?

[@3Dforensics]

This wasn’t my first time trying and I’ve seen what can be done using other modes. If using Lidar only, you will get a lot of noise. Look at other apps and you’ll find a similar thing. Also, if you check out some of my other videos, you’ll see some excellent models created by just walking around. I’ll be doing more of these comparison videos in the future so hopefully you will prefer another one.

[@novarovar2453]

How much the price of IPhone 12 is?

[@3Dforensics]

Not exactly sure but it’s over $1k USD.

[@BatLapin]

Thank you for video. Could you say why is it not possible to switch to "Video+Fusion" mode? Tryin that I get message "Coming Soon". Version is 2.8, using iPad Pro 11 2021.

[@3Dforensics]

I am using a beta version so they are still working on it!

[@anhlangthang3881]

❤️❤️😊🇦🇺

[@yuvrajmakjhaj274]

🇲🇾+𝟏𝟗𝟏𝟖𝟗𝟔𝟎𝟎𝟎𝟓𝟏

[@sigma663]

It is an unfair comparison.

[@3Dforensics]

How so? They are certainly different but can you explain your thinking?