CR-Scan Lizard 3D Scanner vs Photogrammetry - which one wins?
Вставка
- Опубліковано 14 лют 2022
- Following up the recent release video for the CR-Scan Lizard I compare it to Photogrammetry with Meshroom.
The Kickstarter has now finished for this scanner, if you pledged using the channel link, you rock!!!
If you'd like to support the channel please check out the Patreon...
/ makingformotorsport - Авто та транспорт
All KS pledgers :- I am sorry but I have no more hard information than you received in the email from KS. The FB group seems to have the most up to info, I can’t post a link here but you know how to find it 👍
Crossing my fingers as i already had to pay 65% taxes for it.. go socialism!
I got mine today! I mulled the decision over after the KS refund and I did pay for it through pledgebox.
@@MPbdy good man. I am sure that you’ll have lots of fun with it!
@@MakingforMotorsport thank you! I already got a good scan of a part of my engine bay. How do you make sure it is dimensionally correct? I imported into fusion360 and it’s way off
In general to do photogrammetry for a piece like this you need to shoot 150-200 photos including closeups of smaller parts, powder spray the piece and place retroreflective markers on the piece. this will greatly improve the quality of the mesh but your laptop will be in pain from the processing. Meshroom relies on CUDA so you really should have a beefier setup to tackle pieces with this many details.
Keep in mind that the result you got with photogrammetry is not representative for photogrammetry in general. You can have good results too, it just takes more practice and especially time. It's like a skill you have to develop. This is obviously a big advantage for the 3D scanners as you can simply plug and play these things. From what I've seen you took way too little photos. When I scan people's heads the distance between each photo is about 15-20 cm, sometimes even less.
As well as shooting against an unified color, like green screen. Or alternatively you mask your photos to feature only the object to be "scanned" before processing the images, both of these will help a ton. We've scanned hundreds of objects like this and got stellar results. Mind you, focus stacking comes along later when scanning smaller objects at high quality level.
I took about 30+ photos which is probably way too few but I’d be terrified of how long that would take to process 🙈
Like I say, i am sure that there is a load of improvements to be had buuut, if I have to green screen and basically setup a dedicated space and spend days on it then it’s not practical for me. I’d be happy to give another piece of software again if there is a better one…
@@MakingforMotorsport Yea, the setup is quite immense as is the aftercare for the photos to reach proper quality level. It looked like the 3D scanner for your use case is far superior than photogrammetry. Photogrammertry works a lot better for applications that need a texture/material reference as well.
Exactly why I don't use Meshroom
I do photogrammetry of sculptures and a lot of engineering parts. A lot of the time I use it for car parts and "Fan shaped" objects like fan blades or water pump blades. That are really difficult to design without 3d scanning them
With photogrammetry you can get around 0.3mm precission and scan really small objects
It's relatively fast and good if you do it right. You have to spray pain the object just a little form far away to generate detail and a mist to cover the reflections (like, old dirt dust)
You take 50-100-300 photos and conver them
+ better if you scan it with really good light, like in the shadow outdoor on in a cloudy day
With a good PC in 1-2hr you can get a 20 million poligon object created, but I use Reality Capture, wich is like 10x faster than meshroom
I've just done this with my EF civic, to fix the front end. Took a video walkaround of the front using a high fps modern smartphone. Split the video into frames using ffmpeg; the use the standard workflow in VisualSFM to generate a sparse point cloud; then a dense model.
Good idea. Tried it myself, once.
Can't quite figure out why there isn't any photogrammetry software that takes video instead of photos. You instantly get a reference of movement to work with to.
@@FlexDRG Yeah it was a bit of a "Duuuh" moment for me, obviously there will always be junk to clean up, but without a scanner this method's output is very impressive. Last step is to measure min/max dimensions on certain straight edges and hole positions; that way you can bind your messy model to real world units.
@@FlexDRG Meshroom can break videos into frames, and can process as a time series; then reprocess using camera positions to get better fidelity.
@@3dkiwi920 even better, include 3D markers, those help get the original camera position estimation better, and get the scale right. (Apparently), I don't have a printer, so haven't been able to test.
Hi guys, if you are interested in the Kickstarter then please consider using this affiliate link to support the channel :-
creality-cr-scan-lizard.kckb.st/f6f711d9
£242 isnt a lot of money for the scanner.....
Waiting on my Pop2 to arrive and based off your scans I'm now waiting for CR-Lizard to arrive.
Going to do a side by side review
@@nathanpallavicini6687 collecting the full set! 👍
@@MakingforMotorsport Yes, after working in industry for Autodesk, Formlabs, 3D Systems, Markforged I have always needed I mean wanted a scanner but in Australia we are talking AUD 30k +. So been waiting for these solutions to mature after playing with kinnect, Intel L515. Now is the is the time as they are hitting 0.05 reasonably consistently.
And as you mentioned, the more scanning, post processing ect the better and more accurate you're scans will be.
Can the CR Lizard use markers at all on the parts?
Done! I’ve been thinking about getting a 3D scanner for a while but they have been cost prohibitive until now, this looks perfect for my applications.
Appreciate the candid, unaltered, experience being shown. Good result from the Creality. I am going to check that out.
Thank you. This put me at ease over my kickstarter support for my lizard. Cannot wait to start using it and hopefully Mac support comes soon.
i saw your first review of the scanner and went on kickstarter straight up and ordered one 9 min into the super early bird sale. cant wait to receive it. keep up the good video, love it
Then you got a bargain! Hope it works well for 👍
So glad I stumbled upon your channel. Just backed the kickstarter. Keep up the good work!
Cheers bud! Thanks for using the link! 👍
I am looking for a proper link to order one now.
@@rowenbedeau9863 if you’d like to support the channel then please use my affiliate link, if not, a quick search on KS will get you there….
Brilliant video! Even taking into account the comments the time saved is enough for me. Great work!
All the commenters have a point, and as I say in the summary, photogrammetry can do MUCH better than this, but essentially only if I spend so much time on it I make photogrammetry my hobby instead.
I already have too many hobbies, I need something to save me time… 👍
Great video. Small tip: For 3D scanning bolt holes can be difficult to pick up so leaving bolts in them partially unthreaded gives the scanner something much easier to pick up giving you an accurate location for the hole pattern.
Yeah, the holes always struggle, I will may turn up a set of dummy bolts with a threaded and a plain shank to screw in. Hopefully the plain shank sticking up with be picked up better, at least it’ll give the centreline…
You've pretty much made me decide to eventually get this scanner. Seems like the ideal solution for awkward to model parts.
Thanks man … that was an actually helpful video.
Some Harry Hill's TV Burp nostalgia in that intro!
Love Harry Hill! GOAL!
Great video. I've backed the Kickstarter based on your results. More scan videos!
Thanks Daniel… there will be more coming but I’ve got a couple of other vids to do first…
From my limited experience in photogrammetry it tells me that objects that are reflective, even if just slightly so can be a real problem as the surface tends to warp, now there are fixes to this I know (at least when using RealityCapture), but the best result I got was to use photogrammetry in tangent with a laser scanner. That way you will get the surface quality of photogrammetry and the mostly solid result of a laser scanner. I must say though, I really like that handy 3D scanner, seems quite nifty.
Nice vid. the scanner did such a good job.
Nice video, i'd be really interested to see more videos on actually using the scanner, then taking the scans into Fusion to design some bits from. Show us how you take the model, create planes and use the actual scan geometry to model from. Keep up the good work!
I will be integrating this into the workflow for future vids so you’ll see it pop up…
But I recommend looking @Learn Everything About Design YT channel, he has some good videos
Have to say, I discovered your channel a few months back. And I now eagerly await your videos. You've seriously got to get on Instagram though
Cheers bud, plenty planned… just a constant battle to get the time on it… insta??? Maybe….
Not a Fanboy ;) but a couple of pointers what my experience was so far (I also got very terrible first results)
Imho 33 pictures is way too little. I usually go min. 100-150. And for the enginebay it was like 200-250.
When taking the pictures i also start at a bit of a distance (to get reference points) but then 2/3 or more of the pictures are more closeup of the detail i want to capture (boltholes for example).
I also had very good results using the flash on the camera, even if they advise not to use flash...
When in meshroom you should play with the settings for "Feature Extraction" and only calculate the "Structure from Motion"-Node to have a look at the quality of the pointcloud. (This is much faster than running all nodes to the end)
If i get a lot of Artefacs in the pointcloud i increase the "min. input tracklength" in the "SFM"-Node.
The full mesh calculation is for over-night if you dont have access to very good new hardware (massive hidden cost) ;)
I have to say the real scan result looks very nice! But i think you could get close enough with photogrammetry with more effort.
But its like everywhere, you get what you pay for =D or pay with your time ;)
Thanks for the video!
Cheers bud, how long does it take you to process one of these? Do you have a very very VERY powerful PC?
Agreed with the number of pictures. I have never managed to get fantastic results using Meshroom, but I have certainly got way better results than those shown here. Like you say, I will normally run at least 100 photos for a detailed item. I have a GTX 1050TI with 12 CPU cores and the processing of a detailed scan is an overnight activity.
How does Meshroom handle factual size of the objects? There is no inherent scale reference in your photos, right?
i agree with op the uploader didn't use enough pics
@@MakingforMotorsport I wish i still had access to the Workstation from my old work =D
Checked the most recent scan: On a Laptop with an i5 11400H 16gb RAM RTX 3060 with 110 pictures (10MP) it was 44 min to a mesh (until SFM: 14min).
On the older i7 3610qm 12gb RAM GTX570 egpu i feel it would have taken double or triple.
Key for me is to compute/optimize SFM until i am happy with that, as the meshing itself takes ages.
I would say overall its always a 3-8 hours ordeal depending on how good/bad the input is and how many optimizing iterations it takes.
From wanting a scan to having a useable scan Its a f***in process for shure =D
Scan lizard turned out great, thanks for posting! Photogrammetry should've turned out better though. Meshroom is free but doesn't generate a great mesh, recap gives the best results from my tests. Shiny areas need to be dulled down even more or the surface will get bumpy, could use a better dulling spray since there's still a lot of shine. Also, your body is casting shadows as you move around so photogrammetry software will get confused even with slight variations. Putting your camera in full manual mode would help too since the more consistent the photos, the better.
Quite informative.
Cheers bud! 👍
awsome vid man. this the same dif fim using in my AWD rx7 conversion
Yeah the 188! Inside lurks a 4 plate conversion on the standard ramp angles… just need another Beemer now to fit it to!
@@MakingforMotorsport for sure, Im using an E53 front diff. ZF8HP70X from an E70 X5. Should take a bit of a beating. I welded my diff. I HATE welded diffs but cant affford a proper LSD i might try one of them Cheap LSD blocks. Reports that they do work
@@BEYTEK welding a diff ain’t great but needs must, 1 tyre fire ain’t getting you anywhere… I have heard good things (well, positive anyway) about the clamps, they add just enough preload… not a full plate diff for sure but do make a difference….
@@MakingforMotorsport i think so man. Preload enough to turn the steering wheel or turn out junctions with out slapping about haha
While I'm a photogrammetry fan boy, I have to give it to the scanner in this case.
It would take A LOT OF WORK to get similar quality output from photogrammetry even if an experienced photogrammetry user did the work.
Given the equipment cost difference between CR-Scan Lizard and photogrammetry, the time needed for processing and skill level required for a good scan, it's a no brainer.
One other thing that wasn't mentioned in the video is that the CR-Scan Lizard 3d scan is already to scale, while the photogrammetry model is not. you'd have to do extra work to add scale to the photogrammetry model.
Adrian, absolutely my thoughts in this case, it’s a case by case I think. Photogrammetry is a tool and it will work with some cases better than others 👍
well, this pretty much convinced me to splurge on a scanner as well, shows how useful itll be when making a model of flanges for welding later.
Same. Went straight on the Kickstarter after watching this.
The possibilities are endless.
Hi, thanks for a very interesting channel. I do scanning as part of my job, typically I scan mining equipment engines. My current scanner is a GoScan. I'd really appreciate your experience with the lizard working in tight spaces eg down the side of an engine in an engine bay. What I find is the GoScan, as good as it is, is just too big to get into maybe 20% of the places I have to capture, typically around the exhaust manifold. Having to be 400mm away from the object makes life really hard. I'd love to know how the lizard performs getting in and around somewhere where space to move is confined.
Yet another awesome video, this scanning tool is on my wish list. Your mesh room results matched my own. Sadly it's just not a precision tool. Better with more simplified geometry or slab sided objects but like you saw it's more so a suggestion of dimensions vs anything concrete.
I think there are better options and methods out there for photogrammetry but for me going forwards it’s gonna be scanning for these objects…
@@MakingforMotorsport Oh yes it's a far superior technique. Thank you for this side by side comparison. Really drove home the benefits of adding this tool to ones collection.
When I think about it I actually think the CR scan will save individuals money in the long run. For instance when developing my intake turn signals for my MX5 I used meshroom but since it wasn't very detailed it required me to do many iterative test prints to get the measurements just right. I think when I tallied up the cost of development in materials it was somewhere around $80 all said and done if I recall. Honestly a lot of that went to user error since I was and still am relatively new to 3d printing but regardless it'd require multiple prints which all add up. I think if I used a tool like this I would have had it dialed in, in a fraction of the time and cost.
@@MakingforMotorsportImo photogrametry is hard, in order to get really accurate and reliable scans I cover objects with chalk spray and then add small black dots using water soluable wet paint and toothbrush. Such treatment creates ideal surface for optical measurment, scanners using structured light benefit from matt surface so using scanning spray will give you better results. For photogrametry you can give a try to 3df zephyr, I think you will get better results
Amazing.
I am new to the tech, but I am constantly impressed by the tech…
Do you know if there is a Mac software for the Photogrammetry? MeshForm is only Windows/Linux. Thanks!
Would mounting a diffused light ring around the scanner help you obtain a better scan or does it utilize/project an unseen wavelength of light to see the object?
I don’t actually know the tech, but lighting for the scanner doesn’t seem to make much difference…
For anyone using photogrammetry, use 3df Zepher instead of Meshroom. It is much easier and faster. I've used it for control arms, light housings, and even houses (with drone photos). The only tricky part is making sure the final model is scaled correctly and imported into CAD with the correct units.
50 photos limit for a free version.
Hey, I liked your video! I just wanted to comment and leave some helpful advice for better photogrammetry results, should you want to tackle it again. Because to be honest, unless you buy one of the crazy expensive handheld scanners, you will get the absolute best results with photogrammetry.
There are 3 major things you can adjust to get superb results:
1. Image noise greatly affects your results. So making sure you are working on low (preferbly lowest) ISO is important.
2. Depth of field plays a major role, too. So ensuring you have a decent amount of dof is crucial... hence set your f-stop to at least f8 or preferably f16.
As you can tell both of these adjustments will mean that your image is gonna be quite dark. To compensate for that use a longer exposure or better yet, use a flash light.
3. This one is probably overkill but the absolute best results are achieved by cross-polarized lighting. Cross-polarized lighting is done to remove reflections of your main light on glossy surfaces. It's less of an issue on matte surfaces but if your surface reflects your main lights and you take photos from different angles those highlights will obviously shift on the scanned surfaces. Cross polarized lighting works in the following way. You attach a polarization filter onto your lens (usually used in photography to make the sky appear darker or for reducing reflection on glass or water). By rotating the lens filter you basically block the wave length of the incoming light. And now for the "cross" part. Your main lights also get a polarization filter (a polarization film/foil) in front of them. By adding them to you light and rotating the polarization filter on your lens you can entirely remove any reflections of your lights on the scanned object. The resulting photos look absolutely freaky because they are fully flat/diffuse. This technique is especially used for photogrammetry that requires you to extract flat textures as well. But it also leads to way better results because the flatter the photo, the easier it is for the software to discern and recognize unique reference points.
I made did a project at home for achieving high end photogrammetry scans (including the cross polarized setup) and here's two results:
andreas-mass.com/en/portfolio/3dscan_photogrammetry/
andreas-mass.com/en/portfolio/3dscan_mandelhoernchen/
The pastry one is especially a good example since it was very glossy to begin with (thanks to the syrup).
Some fantastic tips there and looking at your results there is some fantastic models! There have been lots of comments (as I knew there would be) and I absolutely will give photogrammetry another go, incorporating as many of the tips as I can.
That said, I don’t have the time or the space to do the setup that some people are saying is needed to get the best out of photogrammetry and I’d reckon that most people watching don’t either… but for some applications, photogrammetry is definitely the best option….
For all that trouble I doubt scanning it is that big of a difference. I'm not printing out museum sculptures. Even as a pro photographer I wouldn't want to sit there and take 200 pictures of something. Thankfully I'm a good designer and I haven't needed to do that.
In another year hand held scanners will be more affordable. I can already see things heading in that direction. Companies will want to compete with this new scanner.
Put the camera on a rolling mike stand, make a circle to follow around to maintain focus, and put a barrier behind the shots to minimize targets.
Which file format is the best to use in fusion? Non stl i think?
Thanks for the vid! Could you please do a second part comparing the file size? I'm interested in creality lizard, currently using a lot photogrammetry and files and stupidly heavy!
Hmmm… my file size on the scanning is 300Mb plus for each scan, and the exported OBJs are huge aswell with 2 or 3 million triangles. When processing the software was using over 11Gb of RAM. So if you want less data I am not sure scanning is the answer 😂
@@MakingforMotorsport I did a gearbox yesterday 200 photos with Canon. Took 5 hours to calculate with quadro p1000 and 32gb ram. The file is 1.56gb and the mesh is ugly :(
I'm definitely buying a scanner xD
Thanks for this video
@@BackyardGarage999 yeah, that’s what I am afraid of doing, sinking lots of time, and getting no further… at least with the scanner, you see it’s not working, stop, change and go again…
Can you tell us the size of the 12v barrel connector? 5.5*2.5? or 5.5*2.1? or something else?
Great info there 👍🏼, what happened to part 2 throttle bodies on festa?
Yeah… it’s coming but the last scanner video did soooo well I kind had to follow it up… but Project Siesta is continuing in the background… hopefully receiving some bits tomorrow!
Another great video thanks. Time for me to pay the extra to get my scanner sent DHL in 7 days rather than wait the 35 days shipping.
Oh man, that would be a long 35days waiting…. Hope you enjoy it!
It would be nice to compare these results to the new LIDAR on the iPhone / iPad
I've seen a few videos on iPhone/iPad lidar vs consumer scanners and it doesn't do quite so well on picking up fine details. Does look quite good when textures are applied at the same time but the detail in the mesh just isn't there.
when you scan it do you get the actual size? or do you need to manually mesure 2 points and then scale it?
With the scan, it’s to size, as accurate as the scanner can do. Photogrammetry needs to be scaled.
there are some ToF 3D depth sensors available for cheap like a Vzense DCAM710 (300 bucks) but i havent found a software yet that allows for 3d model scanning. i think ToF has more potential than structured light scanning, especially with a high resolution ToF camera.
I think the photogrammetry result would be improved by taking more photos at angles / heights in between those used, and (not sure if you did this) different distances from the object to be scanned. Would be interesting to see a comparison using a larger set of photos and showing the difference between using say only one distance from the object vs several, pictures taken at every 5 degrees of rotation vs 10, etc.
Great video. I have noticed that you are fairly soft spoken, like you're trying to not wake the baby up as compared to other youtube videos that I watch. In future videos, crank up that volume for us. Can't wait for the next video!
Cheers bud, sound is so tricky to get right and I manage to mess it up good and proper on occasion!
I used photogrammetry to model a spring seat/shockmount on my volvo from which I grabbed all sorts of measurements: planes, holes, etc. I designed and then 3d printed an adapter which fit perfectly into the part. I really think photogrammetry can give you much better results than what you achieved. You definitely didn't take enough pictures,that result was really bad. Take like 100+ photos with much more gradiation between angles. Also there were still quite a lot of shadows on that diff, it could've been a little better lit. I used a bunch of LED shoplights. Every shot has to be crisp too. Setting the aperture to F11 or more for a wide depth of field is key which means you need even more light! Processing time was very long but the computer does the work and I was free to work on other things.
All of the tutorials on UA-cam were pretty bad from what I remember.
Yeah, I had my lens at F7.1, stopped down anymore I just ran into noise from high ISO! And I already had all the lights on it! 🤦♂️. I will probably give it another go but as an experiment as if an object is scannable I’ll probably just use that!
@@MakingforMotorsport Yup, the iso noise will hurt too. I had a TON of light on this thing and it worked very well, although, I went through the whole process like 3 times in order to get good results. If I had had access to a 3D scanner at any point, I would’ve just used the the scanner haha.
Unless it's a huge difference I would never want to go through all that trouble when I could just scan something. I'm a busy person.
@@dangerous8333 Yeah well if you have a scanner, then that is going to be the better option. The benefit of photogrammetry is that it’s accessible for people like me who already have access to a nice camera.
Try a cross polarization lens filter with a high power ring light in combination with existing lights, a tripod for your camera, black background and mount on a turntable with a welded pipe or custom made mount. 150-200 photos from different elevations. I think you will be amazed. Try Reality Capture as your software. I have been in the field for 7 years.
Wonderful
P.S. Nice meal
That’s a genuine fresh made butchers sausage roll! Quality! 😂👍
Add more photos for just the sections with low detail and run Mesh again to verify if photogrammetry /can/ do the job as well as the scanner. Contrast and a smooth/solid background can also help improve results.
I am sure that Meshroom with the right amount of time and pictures can do a good right. But spending 5times the time on it just isn’t a solution for me. I will revisit photogrammetry, but when it’s a better fit for the thing being scanned.
Tried photogrammetry a while back and was severly disapointed. Now I am wainting for my 3D scanner to get here 😉
Good man, hope it works as well for you as this has for me!
I'm wondering if it will work with Meshlab. That's an insane scanning software. Thought Fusion was tough, but that thing, is a real brain buster.
It outputs Stl and obj so I reckon you could import straight in… I have never used it but it looks similar to meshmixer
Have you been using the scanner much since this video? I backed on KS and I'm trying to decide if I should REback on pledgebox
I have used it abit, doing another video now. I am planning to compare it to a POP2 soon, we’ll see which is better. If you have the option to buy for £300ish then i think it’s great value compared to twice that for the Pop2. The alternative being the RevoPoint Mini on KS at the moment which does look impressive but has a narrower usage case.
@@MakingforMotorsport awesome! When is it due to come out? I'm really tempted but I'll be using it for 99% car stuff so your videos are perfect examples!
@@jonwebb9261 the RevoPoint Mini is on KickStarter right now for a bargain price but will take some months to ship.
To be honest it’s new to me so I am not an expert, but the Lizard is a good useful tool, I look forward to comparing as many as possible 👍
@@MakingforMotorsport yeah that's the other tricky but thing! Having 2 options with not much to compare. I trust revopoint more as a company but I'd rather have something sooner🤔
@@jonwebb9261 there’s an awful lot of bollocks in this field at the moment, with a lot of amateur people chasing 0.1’s of a mm.
The simple fact is that, outside of professional industry, virtually all of these scanners are far better than a home gamer needs, and any accuracy problems can be fixed with 10mins and a digital calliper.
If you want to do minatures or create digital assets or sculptures, then that’s a different deal, but for our kind of Reverse Engineering or 3D modelling applications, even the RevoPoint Mini goes beyond what we need.
And more accuracy and resolution means more data which means larger files which means at some point you will likely have to sacrifice that detail to get a useable file.
We aren’t going metrology. Or at least I am not. So far, the Lizard has met my needs. Would a Pop2 do better? Maybe but not by enough that i see it in the scans I see on the internet. Could I even use that extra data? Probably not.
I wish I'd got in on the scanner when they were doing the first 300 for like £160, hopefully in the near future after release they will come down in price used and I can pick one up for ~£200
There is likely to be a few RevoPoint Pop1 going 2nd hand soon when the Pop2 gets delivered…
That scanner model was really impressive. I'm sure the photogrammetry could have been better with more pictures. Something I've never seen is recording a film of the object at a slow frame rate, say 5fps. Then exporting the frames as pictures. This would enable the user to just walk around the object in a similar way to the scanner.
Videos aren't really good because they have worse compression, artifacts caused by movement, etc.
Sir what software did you used for MultiRubild and Fuse at 7:02
The software that came with the hardware. It’s now been superseded so this video is out of date.
the lizard looks like a great scanner if the tracking issues are reduced. its 0.05mm accuracy is better than i thought it would be.
I'm really wondering how this compares to the Revo point pop. They are similar in price but I think this might be a bit better but it's hard to tell.
I haven’t been hands on with the Pop but my impression is that this is better than the Pop but whether it better than the Pop2??? We’ll see…
I have a Revopoint Pop and it's as good as this. Accuracy is as good as my ability to measure. The software does a great job of combining separate scans too.
@@ianproudfoot5415 I've watched some people use it on discord. The software on this seems to be a bit easier to use. Does that sound accurate? Maybe Making for Motorsports.just made it look easy lol. I didn't realize there was a pop 2 already.
@@madeintexas3d442 The RevoPoint Hany Scan and Handy Studio software looks very basic on the surface, but it does do a good job for the beginner. I do wish that they had 'pro' versions with more control over the process.
Don't forget 3D Zephyr. Also free, with a 50 pics restriction. And fast.
I will add it to the list…
Which scanner is better? Revopoint POP 2: Precise 3D scanner with 0.1mm accuracy Creality CR-Scan Lizard capturing fine details of sight I think they have different ways of working
They are different but I think the performance will be very similar.. unfortunately I do t have a POP2 to compare, but as soon as I can I will do a comparison…. 👍
How'd you get it so fast?
I've seen the specs of this saying you can scan up to 1500mm. Is that a single scan?
I was wondering if I could use this to scan a can in multiple goes maybe.
I don’t really get that, you could scan a lot bigger than that if you have the time, especially in handheld mode….
@@MakingforMotorsport I meant a car back there.. not Can..🙈
I think they might get my dollars soon.
I've done photogrammetry a lot.. and I agree with the comments around.. you need a hell of a lot more pictures.. I run 1000s of them.. and you want a better software.. reality capture is amazing.. and you "pay as you go" Option makes it really cheap.. 300 pictures model would cost around 12$ or so.. can't remember now..
Why was this in my recommendations list if I was looking for Bhagavad Gita audiobook?
It indeed interested me when I was on 3D printing and been looking for 3D scanning a lot, though not ultimately and mortorsport even less… about 1/1000 of the time in the last 4 years or so… though I`ll have a look at CR-Scan Lizard.. pretty cool recommendation. Idk..
🙄😦🤔
Great review and results, I managed to find a lizard on ebay, so took the plunge. Scan results represent good value, the software is something else though.
I must be 20hrs into scanning an audi o1e gearbox, I cant tell you how many times the software has crashed, operating well within the specs of my cad workstation. I dont know if im missing something, but im down to 1x scan per project (sub 4gb each) not holding much luck merging them together given software stability so far.
Sadly I think i’ll have to return it, the software is simply not mature enough for any moderate engineering detail.
Hi there, the Lizard they haven’t been released yet (and there is some legal wrangling going on) so you must have found one of the review units.
I have had the odd crash but nothing that stopped me getting the results.
Try to keep the individual scans below 1000frames, and merge them rather than trying to capture all the data in one go.
Hopefully all the legal issues with Creality will be solved soon and Creality can continue to support and develop it…
Thanks for the reply, yes I really lucked out on the review unit from ebay! My trouble is that with the gearbox being a transaxle, the geometry is cylindrical and i need to capture a full rotation/pass, meaning even on the lowest frame rate, each pass is around 1k frames, anything less than a full rotation and im finding slight elongation occurring in the scans meaning they wont line up. Frustratingly, when aligning and saving the 3x passes, it is just crash after crash, ive not even got to merging.
Smaller parts, ive had good success though. Actual scanning performance is good, I might just export each pass as an stl, then combine in a separate piece of software. I think for peace of mind i’ll do a clean windows install, start again with fresh drivers etc.
Really appreciate the help, your channel is an excellent resource!
You should use reality capture if you want it to be fast and good quality as well. for me I get really good outputs from photogrammetry but you should really have used a lot more pictures hahaha sometimes more pictures even makes it faster since the software can interpet the scene better. Also best to take some closeups aswell since your photos were from quite far away to really capture the detail
I know photogrammetry can do better, but if time is the limiting factor then Meshroom is out…. I will be trying reality capture as that seems much quicker…. But spending 1hr just taking pictures doesn’t sound like fun! 😂
@@MakingforMotorsport well you can get away with like 100-200 pictures. probably will take not that much time at all. afaik you can even combine the lidar point cloud with the pictures so that way you get the best of both worlds.
Probably not much better than Photogrammetry, but you pay for the convenience of the real-time capture feed back.
What we really need is some affordable software to automatically enhance the 3D model.
Where is the festa
Haha! Good question! Taking longer than I hoped! I’ve got some bits in the post 😉
@@MakingforMotorsport nice one 👍
Unfortunately, Meshroom is known to be very slow, but I bet I could get better results with your photos, there are a couple of techniques (as you would guess) to get a decent result, which are very hard to explain in a UA-cam comment, 🤔 I should make a video about that.
Erkan, if you make that video, I will watch it 👍
@@MakingforMotorsport can you share your dataset with Erkan Özgür Yılmaz?
That would be a great way to get a "how to tweak your settings" video.
What do these scanners cost? That’s a seriously impressive result!
Hi John, here is a link to the KickStarter (affiliate link but no extra cost), there are multiple options but lowest cost is $329.
creality-cr-scan-lizard.kckb.st/f6f711d9
@@MakingforMotorsport just bought one, thank you!!
Meshroom is quite tricky to use. I wouldnt recommend it to somebody who has just started with photogrammetry because the result wont be probably very usable and it takes tons of time… Me personally, I did the best results with Reality Capture - in compare to Meshroom super fast (cca 20-30mins) and also their payment model is ok - u pay only for exported stl files (aprox 1 eur).
Reality Capture has come up a few times… I will give it a go!
This is my old photogrammetry made with old Canon 550d: ua-cam.com/video/zK_IxVjN9wE/v-deo.html You can see quality here even if DSLR camera is old. I scanned it for designing body kit purposes. So upper part of the car was not important for me.
fast and accurate photogrammetry needs a really high end pc, but if you can take high res video, film at say 24 frames a second (not sure if it gets any lower), save those frames as image files, mesh room should then have all the "photos" it could ever need.
I have heard completely different answers on this as most videos will have a longer exposure than a picture so it’ll be much less sharp….
I am sure photogrammetry can do great work, but the more I hear the more work and effort it sounds!
@@MakingforMotorsport Yeah, sorry I was thinking of "red digital" or a "Phantom high speed" camera that can do much higher resolution, crazy sharp images at lower speed...for a premium price.
The scan lizard is fairly impressive.
I am constantly impressed with it for the current sub $400 price tag… full price it’s a harder sell without a water tight usage case, but there is no doubt it’s still impressive tech for the money
It seems though that if you were trying to model something much larger… like an entire car… the scanner would be extremely arduous… but the photogrammetry would possibly be more efficient…?
I think that much larger (so inefficient for a scanner) or much smaller (so you can control the area much better - maybe with a turntable setup) might be a sweetspot for photogrammetry…
@@MakingforMotorsport since you’ve done about 100% more scanning than I have, do you think the photogrammetry would work better with a plain background behind the thing you’re trying to model? Not judging in the slightest, but there was a lot of “visual noise” in the view behind the differential. I wonder how the result may have been different if you’d had a flat white sheet behind it?
It’s no problem judge away! 😂 if I removed the background or cropped the videos or did any number of things I am sure I could improve it but this is part of the problem… that is as tidy as my garage gets at the best of times so putting up a load of sheets ain’t really on the cards 🤷♂️
A lot of comments on here are saying that Meshroom is a nightmare to use at the best of time and there are better packages but even then photogrammetry is a skill in its self… like you wouldn’t say you can weld thin ally just because you have an ac TIG welder, this seems like the same deal…
If any 'patch' of the subject has less than 8 clear images with some perspective shift, you will get uncertain results.
Another name for photogrammetry is 'surface from motion'; deriving the curve of a surface by examining how images of the surface change with movement of the camera.
Many images, lots of perspective shift is the bare minimum.
Scanners have to be 'on-site', photogrammetry works with images sent from anywhere...archival footage of lost machines can make accurate replicas.
In my job we use photogrammetry for measuring volumes of stockpiled earth. It does a good job, you wouldn’t use a scanner for that, but then you don’t need millimetre accuracy.
I will be revisiting photogrammetry but for usage case, (mechanical parts, medium size and limited time to spend) I don’t think photogrammetry is the “right” tool… but I am not finished yet!
I don't know what you've done, but I've had better results with Meshroom with only 20 odd photos.
And I'm not a fanboy. I've just pissed about with it a few times. It's also handy for stuff which is outdoors. Not sure how you'd manage that with the scanner.
Neither do i! But nobody seems to be able to tell me other than saying that black isn’t great and to take more photos!! 😂.
I think larger objects outdoor is the sweet spot for photogrammetry, that definitely on the list to give another go!
For photogramety to work you need 600+ images for that size of part
Yeah, I guessed that I needed more photos but it was already a long winded process as it was! I will have another go at some point but on something that is unscannable like a car…
@@MakingforMotorsport Meshroom is very CPU intensive and sometimes also RAM hungry. From my experience if you have a lot of stuff on backgroud and default settings on feature extraction, you are loosing lots of computing power on them, very dark or blown white background is way better and make room for experimenting with high or ultra settings on feature extraction, but my succesfull reconstructions usually takes betweeen 2-300 recognized camera positions and 6+ hours on dual Xeon server with lots of ram anyway. also if you run reconstruction in stages, after Structure from motion part, you can check on your progress, tweak some numbers and create bounding box for Meshing part of scene you are interested in and improve results and save some post processing time
@@DominikKomzik that’s great information thanks… so it still takes 6hrs even with all the tricks and a dual Xeon crazy power?
Were you happy with the sausage roll and chips?
Absolutely…. My body is a temple and I only eat the finest foods… 😂
Great video, but to be honest I kind of wish it didn't exist - I'm before paycheck and by the time I'll have enough money this thing will be impossible to back thru KickStarter. After that it's going to be more expensive and almost impossible to get. Still keep a good work.
Yeah man, it’s difficult. I try to keep the channel to the roots of spending as little money as possible and to be fair, as far as scanners go this about as cheap as it gets, but it’s still a chunk of cash.
That said, the one thing I don’t mind spending money on is tools…
@@MakingforMotorsport Nah, don't beat yourself about it. It's just that situation with this 3d scanner is complicated - it has to be successful, but not too much. If there is not enough interest and sales this product might die and we won't see another consumer grade 3d scanner of this quality for the long time. On the other hand if it flies of the shelves too fast it'll run into supply chain issues, the interest will drop and the same will happen.
For me it just came at a bad time. I have like 4 projects where I run into issues with modelling - compound curves, lack of reference points. This piece of kit would allow me to make massive progress with them. I just hope I'll manage to nab like the cheapest one right after I get paid.
Kickstarted doesn't take the money out till the campaign ends. Which is March 12. So you still might want to get one while you can
I don't get it, why would you scan motors or car parts? 🤔 what is there to make, that's not allready there?
All the bits that go between. If, for example, I want to fit this E36 diff to an E46 car (because reasons) then I can scan both, and understand what’s required from my desk drinking a beer rather than lying on my back sweating my ass off! 😂
@@MakingforMotorsport so it's more about understanding the dimensions of and inside the car and create then a 3d model which fits in? I thought you guys were scanning car things which you want to duplicate, cause prints are lighter or whatever
yoooo cr scan is looking pretty damn good! hopefully that shit is less than a 1000$ lol
It’s pretty good to be honest, I am consistently impressed! (And currently lowest package on the KickStarter is about $330…. )
This is the way to go with photogrammetry :
ua-cam.com/video/U67RJG6DJ_8/v-deo.html
You can have a super clean and detailed render, even better than a 3d scanner, but takes a lot of time ! 😅
Time is exactly what I don’t have! 😂. I am gonna give it another go and try reality capture aswell…
Well, I feel like I've been "BONED"...After completing the Creality Kickstarter and waiting for shipment on the SCAM Lizard, I got a message that Kickstarter took them down because of a lawsuit... I'm losing faith that I'll ever see it! NOW THEY QUIT UPDATING ME!!!
I get that you are feeling frustrated, you are not alone. Unfortunately, legal proceedings usually mean that a Company has to be careful with communications, but the Lizard isn’t a scam, I have used one! It exists!
Make sure you look on the FB groups for the latest as a lot of guys are tracking it closely.
I don’t FB but you’re probably right.
..just more photos and use polarized filter...
for photogrammetry don't you need to take like a 1000 photos, not just couple dozen
Forget MeshRoom.
Try Capturing Reality.
Interesting… I will look into it!
@@MakingforMotorsport Try with the exact same set of photos, and I bet you'll be amazed at the difference.
@@MakingforMotorsport can recommend as well, it is a paid option however. Epic games have bought it, you can get a license on steam as well from what I remember. Worth a look :)
Reality Capture is way faster then Meshroom. But i also have seen that its possible to combine a 3D scan point cloud in Reality Capture with photogrammetry. Which could be a very interesting thing.
Mixing the two techs sound interesting… hopefully you get the best of both rather than the worst! 😂
This would have been very good if you had put any effort into taking photos every 30 degrees at least and also from different heights in similar intervals. What has been seen in many other videos does not portray at all your results. Other than that the subject is good. You should think first and then do this same video second time, just te photogrammetry part, scan you already have.
Thanks for the comment, but I my aim wasn’t to get the absolute best scan on either method, but to compare them both with similar skill levels (low) in each and similar amount of effort on each scan.
I do fully intend to revisit this, but probably with other software people have recommended 👍
Photogrammetry: How many pictures were taken ? How where they taken ? Did it capture enough ? Also no development spray was used here. If I take 10 shit pictures, photogrammetry will not be able to do much. I bet somebody who does a good job with photogrammetry would have just as good of a results as the scanner. It is 100% possible I guarantee it. It just takes more pictuires and better pictures. It would be better to have a good result in both cases first and then comparing the time and money spend to aquire such result
I'm gonna be *that* guy and say that your photos might have not been all taken the right way. You are leaving way too much gap between each photo, and there are areas in the object that might not have been covered well, hence the look of the model.
In the office our rule of thumb for photogrammetry is 10 degrees difference if there is a lot of detail in the object, so that means 36 photos for one revolution, and we go around the object 3 times. one being centered at the object and the camera parallel to the ground, one with the camera higher than the object and looking downwards, and one with the camera lower than the object, looking upwards.
And this is for an object that is simple with no overhangs or geometry that might hide other parts of the object.
Granted, if you have a good quality laser scanner, the laser scanner will result in a higher accuracy scan.
Thanks for the tips, that’s some practical info, I will try to employ next time I do photogrammetry. 👍
Dude, you can't just grab a camera and do photogrammetry, in your time lapse i can see you standing in one of your lights while you take a photo. You haven't adjusted your bounding box to direct where the sparse point cloud should focus on the reconstruction, that's why you have all the rest of your shop reconstructed as well.
And on top of all, how many pictures did you take, I'd guess about 50 from what it looked like?
Add a 0 to that. You'll be surprised.
I respect what you do, but this leans too much towards the scanner..
1- You are using a free software, famous for being abnoxiously slow, against a commercial scanner.. Reality Capture would have taken minutes to process those photos for actual cents..
2- Most tutorials i've seen about photogrammetry emphasize on taking plenty of photos, yet you took about 30? it's like pretending to cook a chicken in 20 seconds..
Putting some effort on the capturing process, taking ~500 pictures of a complex piece such as the differential should get you a fairly detailed model, wich would take Reality Capture less time than it took Meshroom to process those 30 pics.. for a final cost of about 10 dollars..
I'm absolutely interested on getting the scanner, but was hoping to get a fair comparision.. if it was possible to revisit this video based of what i mentioned for a better understanding of what the scanner real advantages are, would totally use your link when purchasing it..
Also, including the result of printing a flange or something similar based off the scan would be the cherry on top..
This is a fair comment if looking to fully exploit the tech, however the point I come back to repeatedly is time, and time is my precious commodity.
Meshroom is held up as a good solution by many, but Reality Capture has been suggested multiple times, and with the low cost/scan it’s worth a go and I will be using it.
I will be integrating this into my workflow soon so you can see the practical use of the models…
So you compared a free photogrammetry program to a $800 scanner? Geez, how much did Creality pay you for this?
Creality didn’t pay me for this, and Meshroom, despite being free, is often held up as a good tool for photogrammetry. I was also fairly clear that I was comparing the amount of effort required vs the results, not the results in isolation.
Photogrammetry is a very powerful tool, but to get accurate results is a time sponge like nothing else I know. 👍
Of course your photogrammetry failed! You need between 100 and 150 photographs. You need to take so many that you are exhausted and bored, and then take some more. I'm sure the scanner is nice, but this was not really a test at all.
But isn’t that the problem..? If it takes so long to create a decent, usable scan then I will just spend 2hrs measuring it and 2hrs modelling…
I should have made it clearer maybe in the video but this wasn’t about “the best possible scan” it was about which is the most useable “tool”, I spent much more time on the photogrammetry than on the scan but the result was no good. If I have to spend even more time it’s probably not a useable tech for me whilst I can get a scanner for less than $400…
@@MakingforMotorsport LOL! I can't argue with that! And I have ordered the scanner (which will be my 4th). I hope it works as well for me as it did for you!
4th scanner!!! Man you are starting a collection! And it sounds like you know what your are doing aswell, (unlike me) so that always helps!!! 😂
@@MakingforMotorsport Another way to think about it is that the first 3 did not really work as well as I had hoped.
You are comparing a 10fps specialized camera to what looks like a few dozen photos taken by hand. The scanner is better for the task but you need A LOT more photos to make it fair. Anyone who knows about photogrammetry may think this is a paid spot where you intentionally showed photogrammetry at its worst. This is especially evident in the exagerrated way you pointed out how long it takes.
I do think 3d scanners are loads better than photogrammetry but taking 33 photos is doing it wrong. I did a quick look and saw people using 50 photos to scan something as simple as a square foot or so rock.
Hmmm, I get what you are saying but I made the point several times that TIME is the issue here.
I know photogrammetry can be better than I showed but if I have to spend multiple hours to get a usable scan then the technology isn’t usable for me.
I cut it in the end but I did film a piece explaining I have kids and a busy job and any scanning tech needs to work when I can use it (in a garage at 9pm at night) Downing tools in the middle of a day to take 150 pictures because the light is right is not something I can do.
I think photogrammetry has its place (I will be trying it again) but not on a part like this…
I'mma be real, I've never, ever gotten good results from meshroom. Datasets I've gotten great results with in Metashape, are unintelligible blobs in meshroom.
Right then, Metashape and Reality Capture are now both on the list!
@@MakingforMotorsport Never used reality capture. But for my purposes, reverse engineering non-pressure bearing gun parts, I've had exceptionally good results with Metashape. In a data set of about 200 photos, I've gotten scans with dimensional accuracy within a thou or less. Pretty crazy considering it's just photogrammetry and I can get any Joe-Bloke with a DSLR to get data for me with a little instruction.
Honestly the key with any photogrammetry software, is to prep the surfaces well, and take a lot of data. Not fast. But very cheap, very utilitarian.
Really bad idea to photogrammetry pure metal chunk, usually photogrammetry is more accurate and can copy textures
Well that’s why I tried it, to see if Photogrammetry would work for my usage cases and the time I have to spend on it… we know the answer!
@Making for Motorsport , I'd say that creality released a game changing scanner, as a nice photogrammetry setup is 1k$+, even a noobie setup is around 500$ to get stable result, but now here is creality lizard with most affordable set up, nobrainer
Not a fanboy, but a cheapskate.
Meshroom has the incremental scan pipeline.
This lets you take a series of photos, and it builds a sparse pointcloud as you go.
Edit your sparse pointcloud to get a better dense point cloud and model.
I did photogrammetry 6 years ago that looked WAY better than yours.
Like I said in the video, I know photogrammetry can get better results. And I also said that time is my precious commodity so if I have to spend hours and hours and hours to get a good scan with photogrammetry then it’s not any use for me…
Worst comparison ever. how can someone who have never done photogrammetry and has no idea how to do it try to doit to compera it to a scanner. and even with the scanner you had to flip it over several times, coat it, re-scan because it still failed? just a shameless add for the lizzard
Okay, I understand your point but I dont understand why you can’t compare them, especially from a newb point of view, which I make it very clear I am, at both techniques.
In the video I also make it clear time is the main consideration, given enough time you can measure and 3D model anything, photogrammetry and scanning, are really useful tools when they save time or are more accurate, if you are doing it for it’s own sake then it’s a hobby not a tool (that’s my outlook anyway).
But a question, you mention flipping over, I was under the impression that with photogrammetry I cannot move the subject unless I blank out the background? Am I wrong here?
You did a bad job at the photogrammetry sorry to say. But if you can afford a scanner, that is offcourse better.
I don’t doubt it… I am sure I could get better photogrammetry results but if I have to spend hours and hours doing it I may aswell measure and model it 🤷♂️
*Are these BS hit pieces made by 3d scanner companies?*
You chose the absolute most rubbish workflow, then claim that the result is useless for the photogrammetry.
While I have seen MICRON LEVEL PHOTOGRAMMETRY SCANS WITH MY OWN EYES.
ROFL. 🤣
I choice Meshmixer as it had been suggested it was brilliant.
Please suggest to me a better workflow and I will try it…
You are not an expert at Photogrammetry.
Very biased presentation.
Curious. I may be biased, and I am naturally badly placed to objectively judge myself, but I tried not to be and give Meshroom a fair try, now we will probably differ on what “fair try” means.
I will repeat what I said in the video, of course Meshroom can do better, BUT if it going to take me 10times longer to get a comparable scan then it’s not viable when the comparison is a 1hr job with a scanner. If there is no option on a scanner then the logic changes, but I wouldn’t have to spend many hours taking photos and waiting for Meshroom before a $320 scanner (on the Kickstarter) looks like great value…