This is exactly what a good RU-vid channel should be - delivering engaging, non clickbatey content about things viewers didn’t realise were interesting. Thanks Matt.
We might want to mention Matt's Magical "gift for gab": Useful information infused with clever humor, including anecdotes and exaggerated visual cues, which aids the viewers in being better informed about 3D scanner features (along with making purchasing decisions). Yes folks, it's not "what you say"...it's "how you say it" that makes the difference for it to have any meaning. Preposterous comparisons can be effective in audience recall; the lesson here centers around "in all practicality".
I work in one of those professional 3D scanner companies in an engineering position. You are the first hobbyist on RU-vid that I've seen who said "professional tools are obviously much better but probably not worth it for hobbyists", instead of the "iphone does the same or better, don't bother with professional 3D scanners" which seems to be the consensus. Thank you for being objective and actually taking the time and putting in the effort to form an educated opinion instead of spewing clickbaity garbage.
Revopoint scanners are ok but not good. There is a difference between high end scanners and hobby scanners for sure. Most of it is the software. Einstar has really good software because it uses the software of higher end scanners from the same manufacturer. That said I ordered a Creality Raptor this week. First metrology blue laser scanner for a affordable price. Now I own a Creality Raptor, Einstar, Einscan-SP and Revopoint mini. 😂
The shop I'm at got an EinScan for dirt cheap like 2500$ used. To be honest will never mess with a cheap scanner again. Mainly because the details are unmatched and the fact it can scan colors and even can pick up the masking marks from a paint job from 20 years ago.
@@hairyneil it's fantastic the smallest part We have scanned was about 2" in diameter. The scan came out amazing. In fact the Ein Scan we have was used to scan a person's face with precise accuracy. To give an idea it could do eyelashes with really good details.
@@RC-fp1tl my boss bought it from a multi million dollar company who offered it to us because their company had just got the latest and greatest Ein Scanner. We do some machining and machine repair for them every so often so that's how they knew about us.
One thing that can make your Revopoint work a *lot* better is getting one of those stabilized selfie-stick things. Once I started using one of those, I discovered that the quality of my scans improved dramatically.
4:56 Put a few markers on the extremes of the object, then scale it in software after scanning. You can even average your multiple scaling references to get sub polygon accuracy.
@@rodrigomartinelli741 I wouldn't trust the rulers to be come out without being garbled, me thinks the markers at certain points and manual measuring them would be best
A couple of things that I have considered doing when I get around to scanning something. Make scale markers. These are things that will show up in the scan that are of a precise known size, usually a small cube and a square frame you can place on or around the object. Set up a frame of some sort around and over the object that you can use to steady your hand while scanning. The rafters in a garage might do the trick if you tie a line that you can hold with the scanning hand so you can pan in a stable arc. Turn tables are a must.
Hey man, great videos! I'm a geomatics engineer with a masters degree in optical metrology, specializing in digital imaging systems. This topic is RIGHT in my wheel house! First, great summary, there's very little I would add in the broad strokes. On the topic of photogrammety, I have mixed feelings. For your use case, its would be _possible_ to get sub-mm 3D models, but I doubt it would be worth your time. To do so, you would need a decent camera with a fixed focal length (i.e. at least a manual focus, no auto focus), you would need to calibrate the camera (it's possible to do that yourself, but it is very finicky, and easy to trick yourself you did a good job), and you would need to some kind of reliable scale reference. For us geos, we would probably use a total station to survey in a couple targets at the extreme extents of the measurement volume. For better results you your scanners, you probably want more "registration features," which could be those white dots, but I personally have used ping ping balls with great success. They're light-weight, highly spherical, a consistent radius, and if you hit them with a bit of matte white, also have consistent illumination across their surface. You could probably hot glue, or even super-glue the on the surface of the engine for you models, then just knock them off when your done. Just be careful not to deform their shape when you glue them, neither through heat, chemistry nor pressure. Also, **until you need to flip you engine over**, adding a nice checker-board pattern to your work bench would greatly improve your feature tracking, with no need for the ping pong balls. It's really important that during a scan, the subject (i.e. the engine) does not move relative to your registration targets (i.e. the check-board or the ping pong balls). You could also put the two together, use the pattern on the bench to scan in each orientation, and use the ping pong balls to register each scan together after you flip it over. Let me know if you have any quetsions. BTW, if you care to read it, I could link you my masters thesis, which features the ping pong balls in chapter 3
Ping pong balls sound quite large in diameter, how many would you use on an engine? Do you mean using just a few as "general reference points" to help with the tracking, with no use on the actual surface data?
@@HyperSculptor The answer, like all answers in engineering is "it depends". The purpose of the spherical targets in my example is to allow you to reliably register two or more different scans of your subject, even when the subject has otherwise changed orientation relative to its surroundings. Put another way, the purpose of the white dots is mostly to help the scanner track its own position _during_ the scan, the purpose of the spherical targets (which could be ping pong balls, but could be other things) is to allow the user to reliably match the subject's surface _after_ the scans are done. The spherical targets would probably _also_ help the scanner track its own position, but that's not why I suggested them. And yes, they would be a bit large, and obviously if your registration targets are blocking significant portions of your subject, that's a problem, but you do want them to be large enough that their surface measurements can be mapped to a sphere with reasonable precision and accuracy. Too small = not enough details = bad sphere model = bad registration.
@@HyperSculptor Oh also, if you're only using the ping pong balls to match subsequent scans, I'd say you want a minimum of 4 visible in every position you're scanning from. So if you're doing two scans, then you only need 4 as long as you make sure you can see them all in both scans.
@@Kruglord thanks for sharing your experience. I actually like the ping ping tek, am going to test it with photogrammetry and see if it helps the software. Anything optical is fascinating, my background is sculpting the human form in clay etc, then integrating digital tools years later. There are great optical techniques, some of them were already used thousands of years ago.
@@Kruglord I’m curious as to your opinion regarding using 3d printed cubes or similar items placed on the scan target. Obviously at bit more difficult to obtain than ping pong balls. Checkerboard table top seems to be a very good recommendation as well. Seems like a frame with a phone holder and skate wheels may be worth experimenting with too…
Happy New Year, Matt! Thanks for all your great content in 2021. Looking forward to another fun, exciting, dangerous year of your projects. All hail the algorithm!
Thanks for the comparison. I've been passively interested in scanning kit car bodies and making form-fitting body panels and I learned quite a bit from your video. It's reassuring to know that I don't need a pro scanner for the kind of project I want to do.
The BMW engine is not going in the Honda. The Honda is getting another Honda engine. It's in the video near the end sitting next to the BMW engine. What's the BMW going into? It's in next week's video! There's also a pretty big hint hidden in this video.
I assume it's going into the tool chest, it's SO unfast to have to roll it around manually. Interested to see what kind of transmission you use to drive the casters.
T-shirt cannons are for my kids to watch. I’ll watch you for real engineering. Thanks for pulling this together, it’s great to learn where the state of the art is with 3-D capture.
Somehow, you're already 3 steps ahead on everything I've attempted to understand. Thanks for the reassuring conclusion on the state of consumer 3D scanners, I was looking for the missing piece of the market for too long... It's looking like the Revopoint Pop 2 might bring the bar up a bit.
Great video! I'm obviously one of the eight people who are very interested in 3D Scanners, but I am also just a tinkerer who doesn't want to pay 10.000 USD or more for a descent scanner. I made about two scans with my Revopoint with one of them being just about usable, but the part was so simple I could have just measured it with calipers and now the scanner is collecting dust. I hope there will be something available soon for around 1.000 USD, that would be great. Please let us know if you find a scanner in that price range.
What a lot of people do not understand is going from Mesh/point cloud to solid model. It’s not just a easy let me scan this and I’ll have a model. I professionally use a Creaform Metrascan 750 for metrology work. We have scanned random things and reverse engineering parts and depending on what we scanned it was a pain or not.
It's a lot of work. For one of my graphics classes I implemented a surface splatting algorithm for point cloud data (basically just surface interpolation from point clouds). It took most of the semester just to recreate the work that other people had already done, and it still wasn't 100% functional. It's obviously not exactly the same thing, but it's pretty similar to how scanners convert the data they capture to models.
@@TacohMann Yea same concept. I scanned my Subaru fog light mount and man even making that a model took forever. Now doing a door cover like Matt was doing wouldn't be too bad. With software I have at work I can pretty much make a surface model in minutes.
For what it's worth, I found this video 10x more fascinating than a t-shirt cannon. Have you put any thought/research into Kinect-based solutions? I know those have been around for a decade now and are likely very far from state-of-the-art but this seems like a problem that is 10% hardware 90% software/algorithm, and I'm curious how the Kinect's community solutions compare to commercial ones.
@@lesdmark Yeah the hardware has a focusing range for the structured light of maybe between 60cm and 5m, with the golden range being preferably far enough from these extremes. You can't even get it closer to the subject to get that much more detail in.
I use an Xbox One Kinect and tbqh its not much better than the Revopoint. The software is probably a bit better on the Kinect tho since it was made by Microsoft. That and Kinect 2s can be had for next to nothing used.
FWIW with some surface treatment (developer spray, baby powder, etc) and a decent SLR you can get more detail out of photogrammetry than most pro scanners. Dimensional accuracy depends on the quality of your scale reference but I get very accurate results when I compare caliper measurements to scan dimensions. The point here is to say, there are a lot of middle ground options between phone scan and pro scan tool. Quality becomes less of a differentiating factor and its more about how much effort and time it takes to get your scan result. Processing images in good software like reality capture (expensive) requires a decently powerful computer(also expensive) and a lot of time.
The software is what makes the difference in hand scanning, or using photogrammetry. Neither can ever replace the inherent precision of terrestrial lasers, however they can get close enough for many applications.
If you're scanning an old Jag you might as well add the serial number to the file description. I've worked on a few old Jags and they are as one off as any purpose built race car I've worked on. I helped on repairing a damaged passenger door hinge and latch on an XK120 that was in an accident. We had access to two cars the same year and found there are very few interchangeable parts, even the doors would not fit each other. The hinges looked liked they were flame cut and all three were made to different shapes. Inside the door skin is a wood frame and even that was made differently across the three cars. I'm not saying the cars weren't well made they were just repeatability wasn't highly sought after.
"There will be like 8 people who watch this video", has 7.7K likes. Don't underestimate the combination of engineers, fabricators, and 3d printing geeks.
Photogrammetry is being eclipsed by LiDAR. However, using it from photographs, the images need to be taken from a calibrated camera, and lens distortions mapped and corrected for, or the "sensors" (digital cameras on the phone) distortions calibrated. They can be very accurate when used correctly, but any distortions that are not calibrated out, it doesn't take a lot of distortion to create large measurement errors, especially in the vertical plane (relative to the position of the camera..... if the camera is moving around, the vertical plane is also moving around with it and that error gets distributed ALL over the place). (I was a photogrammetrist (aerial photography, surface mapping) in a previous lifetime, btw). I love your channel.
In my experience using both I've found that LIDAR is great for efficient large scale mapping but can't capture small objects in a meaningful capacity. The good thing about photogrammetry is that size is all relative so as long as the camera can see it pretty much anything is fair game.
I bought an Ein Scan 3 years ago for $10,000 with the Reverse engineering package. It's fairly accurate if you use it correctly. I can get hole locations to within .002", sometimes better with it. I've designed a couple parts with it. My favorite was adapting a brake caliper to a different model motorcycle. It came out flawless.
I appreciate your content so very much. I find your videos so compelling, not because they are comedic genius but because they are so information dense. You give info how you got the info why you got the info make a quip that makes me chuckle repeat like 50 times and the video is less than 20 minutes long. Any other youtuber conveying as much info as you do would take 45 minutes and would need a part 2.
Dang I was hoping the news would be better on this. If you can get away with modeling 1 face at a time what I do is take a picture with a ruler laying on the part you need a model of. Then in solidworks (and other CAD software too, Im sure) you can insert a picture into a sketch. After using the ruler to set the scale you can just trace the picture into the sketch. As long as the part isnt too big (or camera too close) to get perspective errors this works incredibly well. But isnt 3D 😔. Very quick and simple for 2D sketches though!
"The revopoint struggles with dark and reflective surfaces" Thank goodness for affordable sublimating 3D scanning sprays! Not saying they'll make the Revopoint etc. as good as pro gear, but they certainly help with scanning issues like that, and leave little to no residue after sublimating.
I've used Heges before, just to add and FYI, it also has ipad-via-wifi-to-remote-capabilities built in. Thanks for making this video, very comprehensive, been here before and back then I was looking for a video like this but couldn't find it. Now I know I don't/do need to spend many bucks to buy a hardware scanner based on the project :)
If they were a bit cheaper, there might be a business model around hiring out the high quality scanners. A bit like people can hire expensive tools they only nee once a year, or expensive camera lenses or things like that.
@Romulus III Yeah just pop the BMW 1000 RR engine under your arm and walk into the local scan shop :-) I can just about lift one on a good day and maybe pivot 20 degrees then quickly put it down !!!
I looked into this -- it's hilariously expensive. Like $3k/week. Line up all your buddies and do a group rent maybe but there will be something you forgot to scan. I ended up going the DIY photogrammetry route.
There is definitely a use case in aerospace. My friend gets sent all over scanning in old parts that no longer have drawings or models. It's also often more handy than CMM for complex 3D prints where programming the CMM could be a pain
We professionally use the Arte Leo scanner. Very expensive (out of the scope of what you mention in the video (garage work) But gets the job done without problems. With a bit of experience, it never loses track of where you are, and even when it does it easily finds it back soon enough. In software later on you can manually stitch together parts of the same scan, but the "automatic stitching" of the software works so well that we never do it manually. Last but not least we never use markers, and we can even scan black surfaces. Have a look online, borrow one (or ask for a demonstration). It worths the video -review if you have the chance.
I wrote stereo stitching in college. You have a mill. Get two high quality cameras, mount them as far apart as is practical, load one of N many apps that perform stitching - I wrote one in C if you hate yourself, and go to town. Plenty produce ply outputs, some are fancy and can stitch a room. If you want help picking, let me know. You can have as much resolution as you please. Fun fact. The woman that created the Loreal makeup tester app backend was in my class.
Absolutely with you on the +/- 1000 buck scanner option . I often wonder why the price of good quality scanners hasn't dropped in recent years. Loved the vid Cheers
I sell Lavanti3D ProScan Scanners, new to the market and coming in cheaper and BETTER than Einscan Sturctured Light Scanners -dual fields of views and 3x the capture rate. Also comes with free software, none of that subscription stuff... They start at $7500 USD.
I use an app called EM3D with my iPhone. I’ve found the measurements it gives me are accurate. Then again I’m using it for clay sculptures and then retopologising to get a less dense mesh. Works for my needs.
One advantage of using an actual digital camera is that you have a broad choice of lenses and the sensor is ***MUCH*** larger. I used CHDK on a Canon running a custom script to automatically create focus-stacked, exposure-bracked image sets for macro work. It would be trivial to add a rotating platform and Arduino control to the system so I could automatically take 360-degree image sets for use in photogrammetry/etc.
I agree with your point about the scaling problem using photogrammetry but including a ruler in the images resolves this. Would love to see you do a video using meshroom or something similar just to see how it does.
7:37 - "I wish there was a middle ground" - there is, my friend. Just get yourself a used professional scanner from between 10-20 years ago. It wont cost you more than a couple of grand and will give you just as good resolution as the Einscan (although depending on the system you might have to take longer scanning and/or processing/editing). Oh and get yourself a better poly decimator - you can retain a lot of that detail whilst still making the mesh less heavy/dense. At the end of the day though you're better off just using the dense mesh to spline/surface over.
I own and use a Creaform Handyscan 700 for my business on a weekly basis. Having the detail is essential for some projects. This thing will realistically go down to 3-5 thousandths but it was also $50k. I do simplify and reduce my meshes. Depends on the job. Einscan makes nice scanners for the most part. LIke you said, anything under about $5k is a toy/joke including the iPhone Lidar but they do have their uses.
I've used the creaforms for a few years now for work and love them. Much easier to use than our zeiss scanners, but there is a tradeoff there obviously. I also built a scanner with a kinect, but it was clear pretty quickly why that's not used for actual reverse engineering or inspection...
Pop scanner scan can be corrected by pausing scan and hitting back a few times and then hit play again. It will pick back up just before where it fell off.
The Pop 2 looks like a big improvement on the Pop although it looks like it still has issues with certain surfaces but it comes with stickers and you can use a temporary spray on covering to make scanning easier.
Honestly, this sort of content is what I love! I use photogrammetry due to that price difference that you mention, and while the scaling is a big issue, it still worked fantastic for my purposes! I built an entire transmission tunnel and floor for a Nissan Pao to RWD swap in a 13B FC. It took quite awhile, and 3 scans that needed manually aligned. Though, all that effort worked out well once I cut out all 36 panels on a plasma CNC table and it all fit within a 1/10 of an inch in most spaces!
I found out about Meshroom photogrammetry in the comments section on Matt's original scanning video and using a micro 4/3rds camera + prime lens I already had along with a ~$400 ring flash and polarizing film and built a very capable photogrammetry rig for my own wacky automotive project (ZX10R engine augmented Nissan Leaf) -- with the help of a dimensional target, scaling properly wasn't too hard. The bigger issue was the compute time -- my Dell laptop as much as 3 days per scan (400+ photos) to produce a model. The computer is a high performance piece of HW that shouldn't be discounted in photogrammetry costs. Camera + lens + flash $1000, computer $2k+
@@derekyoung4402 it might be worth trying reality capture, I've mucked around with meshroom and never got a result I was happy with but the same photos gave a much better result in reality capture. Just using my galaxy a8 with unknown lens properties
me: "what is a video about 3d scanning doing in my subscriptions?" -about to scroll past when i see the uploader- "oh it's from superfastmatt? in that case i'll watch it; i'm sure it'll be interesting" and it certainly was. i love your videos, you make everything interesting even if i don't particularly care about the subject or when you go into something that's way beyond my knowledge base lol
Obviously this item is the perfect candidate for a tool library, I have borrowed thermal cameras to detect thermal gaps in my house from my local library in Melbourne, Australia.
I think 3d scanning is really great, and it has alot of applications apart from engineering. its nice to see a comparison between different levels of the tech and if they can make the professional ones cheaper, then i might consider getting a scanner some day.
I'm not in the market for a 3d scanner. I am not sure why I am watching this video but here I am. Since I watched it, it was a good video, well paced and gets good information across with good visuals. well done.
There is a middle ground between cheap and expensive scanners - borrow, or rent an expensive one. Also there are companies that will do the scans for you for quite reasonable prices. So there are options to get good scans without spending a fortune. Also - do you really need a scanner for couple of scans per year or less?
"Also - do you really need a scanner for couple of scans per year or less?" I take it you haven't seen his (multiple) videos of spending 4 figures on tools to make a 3 figure part instead of buying a commercially available one for slightly more 😆
Problem with renting tools is that doesn't give you much time to learn how to use them. Sure, you can get something done. But it usually takes a few days to really master a tool. Even longer when something heavily software dependent is involved. Nobody wants to rent the tool for longer than they need to just so they can figure it out before using it.
#9 here. I run a 3D printing business and have occasional need for professional level scans. Have found the middle ground: just hire the pro gear! I rented a $50,000 Creaform Handyscan Black Elite, which is basically the same deal as the pro scanner you used but even higher spec, for a couple of days (over a weekend because the 3D community are lovely people). Came with laptop setup, dots, the works. Cost more than a consumer scanner, but it came out of the project budget. If you synchronised with a few people and had a little scan party the cost per item scanned can drop quickly...
I use a 40'000 euro scanner at my job and oh boy is it good. It's a newer design that can track with way less dots (or none) while staying accurate to a tenth of a millimeter. As you might guess, it's used for reverse engineering.
There is a third middle ground, contract a professional company to do your part scan. If they can come out to you it's usually a good deal when you NEED detail, otherwise it's economical to just send it out for 3D scanning.
I was positively surprised when i 3d printed a Touch guard for an ABB RCD and just found the whole cad model (at least the outer shell) on their website free to download. Made modelling around it much, much easier
I have used large expensive commercial scanners like the FARO and even they are not as good as you would hope, models generated through them still require a ton of tweaking to be useful. The lack of progress in this field is one of the reasons we still don't have reliable 100% automated vehicles, they need the lidar scanners and software that can determine what things are to be bullet proof and basically boilerplate level refined.
Never even thought about using the FRONT face scanner for lIDAR - thankyou!, this video now has put me off my ambitions to buy an iphone for its lidar. i'll stick to photogrammetry right now
There is a relation between balance improvement and the facilitation of sensory feedback related to the activation of the plantar cutaneous mechanoreceptors. From a clinical point of view, the application of additional tactile cues may have therapeutic benefits in relation to fall prevention or to improve specific types of chronic pain. Postural Insoles are a great way to help optimize joint alignment. They use wedge and frequency technology to target specialized sensory receptors in the skin of your feet, allowing your brain to create new neural pathways to "activate" otherwise "lazy" muscles.
Having taken quite a wild ride attempting to use photogrammetry (meshroom and some other stuff) to do similar stuff to what you're talking about here, I can confirm it's a real rats nest and the part about scaling -- I've run into exactly that issue with it. I'm very happy you did this because of course I'm now targeted with tons of advertising from time to time -- these 3d scanner solutions and I've been dubious about pretty much all of them. Thanks for taking the time to share this -- love your videos man.
Sounds like a job for Anker: They built a great 3D printer that was faster than many mid range options. I'm sure they could take on mid range scanning.
As always a great video. I love your storytelling with the simple cards with text for the sections in the video. The themed music for certain parts of the video is also amazing.
Glad to see this market overview and to see you've exceeded the 8 views you expected. Seems to me this might make for an interesting side gig. Start up costs under 10K and interesting work as well
Amazing video Matt! Whenever you'll be doing aerodynamic analysis of the Jag can you please make it a little more in depth because that's something i've been interested in doing with my 240 :) i know there's probably only 6 other people who are interested but i'd be very thankful :)
I have a pretty inexpensive (paid 1500$) 3d scanner and I have found I get the best results if I use some powder and also target dots here and there, then I built a lazy Susan that's trying enough to hold something like a motorcycle engine as well as a tripod for the camera. I also use really good lighting so i don't get any shadows. Then I start the scan and turn on the lazy Susan and it slowly rotates
This is why I love your channel, this is exactly the kind of info I was wanting to know but don't have the resources to access. However, I guess that means I'll stick to using an iPhone for scanning since that's what I've got!
