I used photo for a LOT of years in Aerospace to measure large tools and we never used it for 3D mapping. We always used a lot of tie points (fixed points in space larger than the object under inspection that the camera could see from multiple angles. This helps to stitch the photos) We also used scale bars with accurately located targets of a highly accurate distance in multiple locations. You can use reference points on the scale bars as tie points. The bars need to be in location for the whole photo session in all locations with known good distances. LOL I almost cried for you when you said it takes roughly 15 min to process the photos. When we first started using Photo, we might have had 200 points to map and used 3 scale bars in X, Y, Z orientation so the software could establish scale in 3D. Add to that a hundred or so tie in points and we could measure 3 stories tall, 300 feet long and 30 feet wide in space to 0.015" (fifteen one THOUSANDTHS of an inch. Or less than half a MM. Remember this is just points in space, not 3D mapping. We would load the pictures into the computer and start the computing at roughly Noon. THE NEXT MORNING when we came into work the computing MIGHT have been finished! LOL! Leica came out a few times to try to sell us a scanner setup like what you show but on an industrial size and portable . It never worked for us. It was not as accurate as we needed and we did not need to have a detailed 3D map of the surface so we never did any of that. We DID use Laser Trackers to do a lot of work and with multiple stations we could map the points in the tool referenced above to plus or minus 0.007" or less than half the error of photo and we could do it in an 8 hour shift or less start to finish. The laser trackers were easy to use but were not as accurate as Theodolites. Theodolites could measure again to less than half the error of the laser tracker but it took a skilled 3 to 4 man team to do the same job as above. It is all about the job that needs to be done as to what tool to use. We would map the critical points of joining the wings on an aircraft with Theodolites within the same 7 thou and that is a 3D object 250+ feet wide, 40 feet deep and roughly 4 to 6 feet thick in about 5 hours but again, we were not mapping the contour. For mapping contour we would use Laser trackers set to measure a point every time the reflector moved a programmed distance, say every one tenth of an inch. We would sweep the reflector over the surface in a grid pattern then take that resulting hundreds of thousands of points and mesh them into a CAD model of what the part SHOULD be in a perfect world. Uh LOL I guess most of this really does not apply! It is your fault LOL, measurements were my life for over 36 years and I guess I tend to get excited when I see tech being used that I used for so long :D Now that I am retired for 3 years, I miss it and talking about it!!!!!! If you made it this far.. GOD BLESS YOU! now CARRY on :P
Very interesting. I worked at a sawmill about 10 to 15 years ago, and we used high power infrared scanning lasers (USNR/Perceptron LASAR) to scan logs/cants to determine the "best" cuts to optimize for the most revenue and board footage. The software would build a 3D reconstruction, extrapolate the bottom points, then figure out what boards would fit in the volume of wood based on our daily production parameters. Calibrating those things were a pain-- we also used stepped calibration targets where all the dimensions were exactly known. And even then, we still re-measured them with our calipers to be sure. lol
Baby powder works well if you don't want to ruin your object with paint or anything. Apply on it prior to taking picture or scanning. It does increase the quality on shiny or darkish mat texture.
Can anyone explain how to have the final 3D model image of a shiny/ chrome object still have that glossy, shiny but using the chalk spray? I get how you apply it, take photos, create 3D image but do you use post production to add the chrome surface back in after? I’m confused.
yeah, you just use a polarizing filter when doing photogrammetry with cars. Also, shoot the photos in raw or HDR mode (a plus for a 'real camera' over your phone) and Lightroom can remove lens distortion for known lenses.
@@JKC40 My Einscan SE scanner has an HDR mode, but I wonder if I would get more complete captures by placing a polarizing filter over each camera ... worth an experiment
@@BenVanDenBroeck which brand of chalk spray?? ... I have used baby powder, darn cheap, a little messy to apply, but works incredible well, and is easily washed off. I also have used Krylon "dulling spray", it's cheap, but only helps with highly reflective surfaces, with limited results ... while baby powder helps well with both reflective and dark surfaces.
You know.... I woke up this morning not having a clue about this technology and now after your video randomly popping into my suggestions, I really am excited to try this out. Great subject matter!
Tom, I love your channel and thanks for introducing some people to photogrammetry. Those with access to an SLR or at least a tripod with their phone are going to fair a lot better. Keep that Aperature up in the F11 range: the tripod will allow that shutter to be as slow as necessary. Most phones can be set to voice-operation, "cheese, cheese, cheese..." as the tripod gets moved around or the turntable spun with the object. A light dusting of powder or a polarizer or both helps with the glares. Spray chalk is also great for items that are ok to rinse with water. Those reading: please join some of the photogrammetry groups online and get ready for a vertical but satisfying learning curve!
Einscan is based on open source technology. You can actually build your own from scratch. Actually, ideal for a Maker's project. The 'parent' project was called DAVID Laser Scaner, and it had two generations - first one used line laser (you had to buy the laser diode at prohibitively large cost of $5), and used motorized setup and video footage to generate vertex cloud, and second generation included a projector shining stripes onto the object with set of photographs - it increased the cost, but removed the need for perfect mechanical setup for moving laser, or at least prism or mirror. Einscan's dual camera setup is improvement over the original idea, btw which sprung around the time Microsoft's Kinect SDK became available. LED projectors can be had now for less than $15 (they have wild resolution of 320x240, but since all it does is display bars, it should be fine), so it's not that expensive. Sadly DAVID's site was hit with users' data leak at some point, so it had some dark side to it. Edit: I believe it might have been some time since I looked it up. It had became paid, closed software. What a shame :(
David laserscanner had a free version, but it was not updated for a long time and got more and more limited compared to the pro version, then HP bought the whole David project and rebranded the scanner hardware/software as "HP 3D Scan"
There is sophisticated all-in-one package photogrammetry 3D scan solutions out there as well. Although you don't need it and as you said, you can do it on the go, you always get better results with a calibrated well-known setup. By the way, the technique Einscan is using is called "stuctured-light scanning". I think if you compared the two techniques of two machines in the same price range you wouldn't notice much of a difference accuracy-wise.
For me is photogrammetry all the way, its incredibly versatile cheap and has pretty good precision if you know what you are doing, you can scan from insects to mountains!
I was thinking the same. I was not sure untill i read your comment. Aim a jewelry designer and aim looking for a way to scan small objects. I was thinking about a 360c turntable and take picture with my canon D7. ✌🙏⭐👍💜
How many scanners you've used? And which ones? Using photogrammetry (for me) is a major nope. Especially if you have more than a few things to scan. My record with Artec Space Spider is about 70 models per workday.
@@ovDarkness I haven't used many scanners in a work environment, and only demoed an artec eva, you know i can't afford a $20k scanner much less $30k for an artec space spyder. Im talking as a hobbyist, i might not get the same precision as one of those professional scanners but as i said i can scan from insects to mountains, underwater, wherever I can take a camera. Hobby grade 3D scanners aren't that good and still very expensive if you want to get any respectable resolution. Edit: seems the space spider now its under $30k it used to be more expensive, still is a ridiculous price for a clothes iron with 4 modified phone cameras and some LEDs and a mini projector.
@@Vio1007 Sorry im 11 months late YT never alerted me about your comment, i use agisoft photoscan and the camera i use is a canon T3i, when i want to scan very small objects i use a 18-55 lens with a reversing ring, i might add i usually stop down the lens to f11 then remove the lens before reversing it while pressing the show DoF button, that way the lens keeps the aperture while you have it reversed, going to smaller apertures you get more depth of field which is good but you lose a lot of light and some resolution.
I know this is an older video, but I wanted to point out that there are now cellphone apps for photogrammetry that work about as well as shown here, but they generate the model automatically on the cellphone. It's pretty fast, but the detail levels haven't improved much.
Why would anyone give this a thumbs down? I think that this video was very objective and told both sides of the story! What else do you want? He doesn't waste a lot of time but gets right to the point.
Hi Tom, good video but i believe the einscan is also "just" a souped-up camera. They can measure the distances thanks to the projected lines and interpolate thanks to photogrammetry too, but this is not how "real" 3d scanners work. Real 3D scanners use Lasers to measure distances and give you a live preview of the appearing mesh and let you know where you have to go over again (yes you have the scanner hand-held and go around the subject to be scanned). Check out the Artec Eva and Artec Spider for "real" 3D scanners
Yes, yes, technically, the Einscan is a photogrammetry scanner as well, but structured light is as "real" as it gets. It's actually quite similar to what laser-line scanners do.
The Artec Eva / Spider are also structured light scanners with a camera that works on very similar principles to the Einscan, so i don't know what you find that makes them "real". Yes, they have more in-unit processing and they are faster, but a unit that is at least a magnitude more expensive should be... The Einscan Pro2X Plus has a very similar user experience to the Eva (yes, i have used them both)
i brought an einscan a few months back. It works great and turns out 100% usable scans and models. I try the laser scanners and photo ones. They simply didnt work for me. So i say if you needs a working scanner and have the money the Einscan SE is the way to go. Also if you want to scan bigger thing its pretty easy to use a tripod and simply rotate the item for each scan.
@@MrDead1975 Wait an autodesk photogrammetry program that offloads to the cloud? That use to be called 123d catch. 123d used to be absolutely free if you sign up and you don't care about other people stealing your models, when I tried it out I was disappointed, the quality was way worse than I expected and to get a reasonable model you need to take a lot of pictures. Anyways, going back to that recap you're talking about, the site literally says you can have a free TRIAL or use it for free if you are a student or educator. But ONLY IF you are one. The licence clearly says: "it can only be used for educational purposes". The terms of service clearly says: "You must be a Student, Faculty, Qualified Educational Institution or Other Authorized Education User (each an “Education User”) to access or use an Education Offering, including any Software licensed under an Education License Type.". Simple conclusion: it will cost you hundreds of $$$ per year if you're not a student.
Good subject! This is quite a question nowadays. I use photogrammetry for anatomical objects and with a 50 Megapixel camera I reach a level of detail that I could never reach with any other method, so for me and for many of us who already own a good camera, i would definitely conclude that for organically shaped objects photogrammetry is the way to go. Not for reverse engineering purposes perhaps and not for real-time capture needs of course.
@@MakerFarmNLRC is insanely fast compared to the others. But for the hobbyist, the pricing model makes it less appealing. But at work, where I don't pick up the tab, that would be the one to go with. Especially if you have a powerful workstation. What the others would do in 4 hours, I swear RC could do in around 30 minutes. Though I should do another comparison with Agisoft. Supposedly they did a little improvement in speed. And at a one time $180, it isn't as horrible on the pocket book for messing around with at home.
When you say "not for reverse engineering" does that mean you can't get an STL file from your scanning for 3D printing in the end?....surely that is possible with a low resolution iPhone?
Great comparison, i had wondered about the photogrammetry. I have an Einscan SP and it's so amazing to use for copying things that are no longer manufactured, like brackets or clips. Thanks for sharing and as always Keep Building👍
Real scanning and phorogrammetry works different. At work we have a 200.000 euro and 500.000 euro machine for scanning , it is a measuring machine, so it need to be precise. The object gets a lot of stickers (and coded stickers)on the object, a lot of adapters can be used to capture it in the photogrammetry, this is to capture the middle point of a hole(and a lot more). The photogrammetry starts, it will not use the pictures for the mesh, it will take pictures of all stickers(coded stickers), and adapters, and the scale bars. Scale bars are for distance measuring(every year these needs to be calibrated 1000 euro). Now you have a pointcloud of only stickers on the object. On the stickers only you can start measuring if you want. The stickers also good for if camera or objects moves, it will not accept the scan. Then the scanning starts, the scanner calibrates itself and the scanning is very precise. scan the object, every scan will see the stickers, looks in the pointcloud for sticker, finds it, places the scan on the stickers, this is very precise, after all scans are taken, you make the mesh, the stickers are scant, and are higher, but the software removes the stickers, so the mesh is smooth. Auto cut out, to the cad model, make RPS, and you have the real measurement in real life and very precise, and you never can print this quality. The camera lenses can be changed 700 mm (capture area)is the big one, and 400 mm is the smaller one, then you get more precise. We work with blue led light. Newer versions go to 1000 MM and laser light, even bigger and faster. You can buy special can of white spray for the objects, it will dissolve after a few ours without cleaning(expensive). Search on UA-cam for " GOM optical" then you see movies, and how to. GOM has a free version of their software, where you could use the mesh repair smoothing software, if it is useful I do not know, you can import stl repair and export, You cannot import a cad model.
Yep, I am a retired tooling inspector and for many many years used photo, laser trackers and theodolites, Industry is nothing like the toys in the video. Two thumbs up :D
Gotta say, the ability to take a photo of just about anything and print it is more attractive then the ability to take a fantastic photo of whatever will fit on the platform
It is indeed sad that EU approved article 13 :( My country is affected too. Very nice comparison, Thomas! I love how you make your videos! I used in the past the photogrammetry to scan some household stuff. It is good enough to get a rough model of a thing, but you always have to correct the scan in meshmixer or other software. Keep up the good work! :) Cheers!
so Einscan here is not a laser scanner but a stereocamera boosting object resolution by projecting reference patterns? that should be applicable into the smartphone scanning game...
Knowing your love for opensource, I'm supprised you didn't include opensource scanner like the cyclop, freelss or my sardauscan (it is hard to calibrate, but it only cost 30$ is much faster than 30min with decent machine and can be used to automate photogametry process)
I've used some laser-line scanners previously and the results weren't usable at all. But I'll make sure to get my hands on one and see how far I can push it!
Some years ago in 1995 I went to a shopping mall in Cape Town, S.A. and they had a laser scanner that took a multi picture map of you while you sat in a chair (very still for 10 minutes) and then it used the info to make or burn a pseudo hologram of your head in a block of clear acrylic about 40mm X 40mm X 50mm. Would this be a similar process to photogrammetry? The image is very clear with very fine detail.@@MadeWithLayers
Photogrammetry is the best, take hundreds or thousands of pictures to capture some monuments, ruins or castles. I'm a funboy for that. Back in a day there was even a developers who worked on the mobile app and the app was creating a 3d point cloud at the moment when you were taking pictures. But I don't know were they got with it, did they finished it or not. I used agisoft photoscan
This is funny I did my own EinScan vs. photogrammetry comparison. I showed the results in separate videos though, not one real comparison video. I also used a different EinScan model and different photogrammetry software.
Photogrammetry has a leg-up when it comes large subjects, say a large building. COLMAP doesn't even need to have all the photos be from the same camera or the same day! Take your own photos, get other photos from on-line sources or stock photos. Works just fine. Structured light is a great process but it is limited in terms of object size and access to the object. Photogrammetry can work with images of a demolished building, assuming you have lots of photos. Now if only structured light could go back in time.
Been wanting to find out more about this field for awhile, right now we can make objects but not scan objects, same as the printing press came before the camera, I think this industry will bloom in the near future, thanks for the video Tom!
@@ryanmickelwait1521 for photogrammetry. the einscan is using a calibrated light source and any light other that from its source will confuse it. Adding structured light to untextured surfaces is often good enough and always a nicer option than the mess/damage of painting the object.
@@drkastenbrot I guess if you simply use a bath of liquid chalk, then the coating of shiny objects shouldn't be too much of a problem. If you have a chemicaly very stable Objekt, and if the resolution is not of most importance, then you could even try dunkin the object in a CaCO3 solution, let it dry for a fiew minutes, and abuse the fact that the hardening process creates small flakes, and thus creates a perfectly non-smooth surface.
Thomas .. your reviews of the Einscan SE unit convinced me to purchase one .. 4 months later it has already paid for itself through Ebay 3D Printed product sales enabled by this scanner. It does take some time to complete scans of detailed complex parts, but the results are outstanding, and incredibly dimensionally accurate.
@@leecaste Top Secret ... have 100% of that market now, not that you want to do anything evil .. but someone else might read this and upend our many hours of hard work. The cool thing about 3D printing is that ANYONE can be a producer, just find a niche that aligns to your own personal interests and knowledge and social media network .. and you will have fun, have free advertising and make a few bucks.
Thanks for the great info! If you have Photoshop and probably other photo manipulation software you can also adjust the individual pics removing shadows etc then run it through meshroom hope that helps someone :)
I would do that if I knew what to do, but I would use my 15Mp digi camera. Would that be using a video mode to get the multiple photos and at a low resolution to get them all on a memory card?
Love the colorgrade of your video with with that slight white milky effect. Really smooth and comfortable to watch. And props for your english. Wizout ze standard görman accent. ^^
You should take the photos with an actual dslr to get the metadata of the pictures. You can also use photogrammetry with aeral photos like drone footage, which is an advantage. Also the einscan se takes more than 15 min for full color meshes. We have one at my university in trier which i coul use in the past. Still thx thomas for the great vid!
it's nonoptimal to use a phone camera with a rolling shutter for photogrammetry. The phone moves throughout the photo which means geometric contradictions and inaccuracies in the end. A DSLR would be my camera of choice.
Phones also need the OIS turned off, both optical and digital (most can't do this). OIS causes distortions in the image while stabilizing, most people won't notice it, but a computer will. It's most obvious in videos. If left on, many of the images will cause errors in the calculations, and depending on the software, will be left out so you have less photos to do the calculations. Or in others, causes strange results. A cheap point and shoot will do better than the vast majority of smart phones. But like you said, get a DSLR for best results. And if you watch craigslist or ebay, you can get second hand pretty cheap if you keep your eyes open.
The iPhone X and Xs have forward facing 3d scanners. There's some software apps that use the scanner to create a 3D model. Could you do a review of this please?
Maybe you could use a toothbrush to spray paint some black color on the white object. That could give enough texture for the software to find some features.
You can get very low resolution cheap projectors off the internet. Even some portable ones. Projecting a 'calibration pattern' on a feature and textureless object like the angel should dramatically improve the results, no? After all it is kind of copying the Einscan System.
Yes, that should do the trick. Projecting static random noise (that is, every pixel is at least 50% in lightness and no two pixels have the same color next to each other, no animation of any kind) should be pretty good for the object surface. However, you then need to do another scan for the textures, especially if you use handheld camera. The Einscan system can shot an image with full illumination and continue to next angle.
Great discussion, have tried the hand scanners and a lot of photogrammetry, including drones. I guess I would have to assume that my mobile phone and my drone would be my outdoor choices and a hand-held scanner or other would be my inside choice. also if I had a model that I really wanted scanned nicely I guess I would have to make that choice whether a $1,500 us scanner was worth the scan
I find that using a good 20 mp camera where I can manually control all the parameters like f stop, shutter speed, exposure and focus gives way better results than a cell phone camera. I put the camera on a tripod and rotate the subject on a simple turntable with controlled lighting. After much experimenting I have gotten good results with photogrammetry. Have been using free version of 3d zephyr software. You can"t beat the price versus a scanner.
Photogrammetry includes both processes covered here: metrology through clever photo analysis or through some manner of scanning. In fact, the 3D scanner is processing photographs. Some more appropriate terms might be "structure from motion (sfm) versus structured light," or "multiple view geometry versus structured light." Also, don't let your output from Meshroom fool you. It may look great, but to brush aside the greater precision and accuracy of the scanner is a mistake. They are not both capturing your subject "just absolutely perfectly." The scanner's accuracy is probably on the order of hundreds of microns, I'm guessing. Your mesh accuracy from Meshlab was probably one or two orders of magnitude worse, and takes usually hours more, to boot, if you use a proper number of photos and have average hardware.
I tried photogrammetry and did not have much luck and took huge processing time, but using the Xbox 360 Kinect-1 and skanect, it worked shockingly well. Have you tried Skanect? Trial version was a good, enough for me to buy it (130$ I think).
Likely not, because he is not talking about people, but machines and software. If he was indeed talking about a group of people, then without doubt he'd deserve to be demonetized.
hi tom ive not tryed anythink like this for a good few years but i did have amazing results in past with davidscanner back then it just used a webcam logitech 920 pro if i remember right and some paper patterns you printed out and put in a corner, it realy did work well i used it at time to scan a model car for a project, might be worth a look how its moved on
You missed out two categories that would have balanced things somewhat. Scale and flexibility. You can take a camera everywhere, not so much for the Einscan. I have also seen people take scans of large objects where you are limited by size on the Einscan. Still a nice device though.
I collect ancient Greek and Roman coins. Many of us make plaster casts for study and trading by pressing the coin into plasticine clay and filling the mold with liquid plaster. The plaster can be colored if you prefer other than white. This also evens out the color of coins made hard to read due to variations of tone or patina. Have you tried making a plaster cast of a shiny, multicolor or other difficult subject?
In regards to accuracy in photogrammetry, the higher your megapixels the more detail you will capture and in Agisoft for example (around 170$ for the standard version) you get the calibration tool which eliminate distortion if calibrated properly. you can get sub mm results with photogrammetry
Please do not confuse megapixel or resolution with accuracy. They are almost independent, although if you have freaking resolution you can map on top of image by image with more control points. Still if you don't have something really accurate to relate it to begin, it doesn't matter what you resolve. That is why good photogrammetry tools have excellent scales to begin with. Also accuracy is tottaly dependable on part, so yes you may get sub mm but you can also get sub cm or even sub dm.
Luís Filipe Lopes true, but higher resolution, depending you know what you are doing can also yield better accuracy as the solver have more information to work with. It is true though the higher you go the sharper you need your image to be and here lies the experience of photography in general and photogrammetry as well.
@@-Belshazzar- yes, that is partly correct. First resolution, specially without a fringe light projector, can be misinterpreted, and lead to errors in the first place in terms of mapping, and two points of view, will change their mapping because light doesn't offer sharp information to start from. This is why a pattern, is always better, as it is objective. You can also fake it, it will help. Then, the other part is that, even if you don't have resolution, you can still get very accurate results, in the mean value. There is more associated noise, but quality information is not dependable on resolution per se. Of course, if you add precise scales and targets that can be more dependable no matter the light, than you can achieve a much better photogrammetry model to start from, but in here it lies, it's not dependable of resolution of images, but quality of artifacts. Also, if you don't have enough form (and not texture as it's told in this video incorrectly) the photogrammetry system will loose itself very easy. This is why it's not so difficult to scan small parts that can fit inside the sensor scope, but larger parts are worst.
I wonder if a phone will ever come out with built in hardware for scanning. Possibly by replacing the camera LED with laser projector and using that to project grid lines during scanning, along with the software to automate the end to end process of building the model.
@@oudent Interesting. I am planning on buying a Nvidia card to do this but I may go for a slightly cheaper older gently-used one. I've seen the autodesk one produce good results but also completely fail. I'm not too worried that he said "probably" but "may have" would have been better and stating what that hardware was would have been best.
Photogrametry for me. As an analogy, we can look at the time of VHS vs. Beta? VHS has been more popular since it's more practical and more affordable for the common consumer as well as the industry that caters to the VHS or Betamax crowd. For instance, when one gets into a car accident or need that extra information in a pinch, the photogrammetric images that can be taken in their phones is flexible enough that you don't have to worry about when and where to take out that device just to take all those pictures because your good to go everywhere and anywhere you bring your phone. The 3D scanner, you have to pull out of storage just to use it (you pull it out of storage since it's an expensive device and you don't want just anyone to tinkering with it who don't know how to use it). Going a step further, for crime scene investigation photogrammetry may be more preferable since the investigator taking pictures is already knowledgeable enough with the camera not to warrant an extra step to set up a 3D scanner to mess with any evidence.
Thank you for the video, very interesting. I have a few comments though. You are not really comparing photogrammetry to 3d scanning. You are really comparing basic photogrammetry to stereoscopic, structured light photogrammetry. True 3D scanners (mostly LIDAR) directly measure the range from surface to sensor (TOF of the laser to points in the surface), rather than indirectly calculating it (which is why the also can directly scale the objects). In general, scale in photogrammetry is arbitrary or very inexact (even when it successfully reconstructs a proper object), because in pure photogrammetry it has no idea how far you were when taking the images, and stereo differences become very unreliable when distance is above a certain ratio of disparity, and projectors really struggle with distance (which makes the patterns unusable).
You sound like you know a thing or two. Could you help point me in the right direction? I want to scan interiors of buildings AND products. I also have a limited budget so am looking for creative solutions. I see people using structure sensor for it, and then you basically have to take the rough point cloud / meshlab and just use it as a template to rebuild the room in a 3d program. This Einscan seems to be more powerful that the structure sensor, but you don't see any talk of people using it for scanning buildings. Something that would provide a more accurate mesh would make workflow quicker, but I would like to keep it no more than 1500 to buy
Personally I prefer laser scanners. You pretty much need to spray with developer either way, but for laser scanners attached to articulating arms, you don't always needs target stickers. Just one reference point if you plan to splice scan data into 1 mesh.
I'm amazed that 3D scanners are still so ridiculously expensive, and limited in size and versatility. How hard is it to attach a couple of reference markers then operate at a much larger scale? Also, did I miss something or did you not talk about texture capture at all?
@@Martial-Mat Depends on reasonable. Haha I've seen a few I think that are under $1000 but I don't know the quality level. I've only used the EinScan Pro+.
Wow this was very informative, thanks! I haven’t read all the comments but one limitation I see with einscan is the size of the object. I’d like to scan some r/c airplane models that are up two - three meters in length, I think einscan would have an issue??
I love the video Thomas, thank you so much for sharing. I have a quick question. Are there any 3d scanners that can see through hair, fur, or clothing? I know the airports use mm-Wave technology to see through garments, and thermal doesn't really work either. Any ideas?
If you can get such good results with a smartphone camera, what kind of results can you get with a high end camera with the right lense, and more pictures? (Yes, I expect it to take a point away from photogrammetry, in the cost category, but I think it'll add one in the performance category.)
Hello Thomas.Why not use VisualSFM for photogrammetry and MeshLab for proccess the 3d meshes? I m doin my thesis at photogrammetry and made a scanner also, very instersting field.
I would be happy to talk to you about photo. I did it in Aerospace for a great number of years. Real life can sometimes be ...... different than what the books say :P
I can recommend OpenMVS. It scales well to low thousands of images. The reduction in actual manual work is something I'm currently working on, because at the scales I'm aiming for manual work is not really an option. In case anyone here has a camera with an electronic global shutter (fancy video cameras like REDs or black magic and similar usually offer it), I'd be interested in traiding a suitable raw capture for the reconstruction from that capture. Contact me ;)
First off, amazing video, loved the information in it. You should do a video showing different techniques to capture hard to capture objects. You mentioned some that wouldn't hurt the integrity of the original model (talcum powder for example), and that would be interesting to see. One suggestion, that I don't think would affect your rankings or anything, but is important to take into account, is you should have a solid, properly colored background. I see this all the time in people showing off paint jobs on miniatures, or even just attempting to sell them. If you take the pictures with any organic background, it can affect things like lighting, or just in general affect the image. A solid neutral color background could possibly affect the lighting properties, and allow for a "better" scan. Worth an attempt, or even a "how to" or "do/don't" video for scanning.
What I missed in this test, is how much effort is it to create a watertight model for 3D printing, The Einscan does it automatic in the software, using Photogramity you would have to do that yourself, and can take some effort to get rid of all unwanted parts, to create a closed model. Einscan - 5 Photogramity - 2
I use the program Agisoft Metashape for photogrammetric processing pictures and it can cap off models or fill holes. But this program cost money. (179 dollars) I also use (from work) a 50K real laser scanner from creaform. This is a high resolution scanner that scans up to 2 micrometer depth. But you get penniless only by scanning, it needs special measure stickers costing 150 dollars for 200 stickers. and you need a lot of it. I use +/- 100 stickers for a 2 by 2 meter scanning area.
Why in the world would you compare photogrammetry with scanning using a cellphone for the photogrammetry part? That's like comparing bicycles with motorcycles using a single gear bike. Photogrammetry using an actual DSLR (even a $400 mirrorless one) is going to give you results far and beyond what you're getting with that cellphone. This comparison deserves a redo with a proper camera.
I'm not sure this is a fair comparison. Some hastily snapped phone pics isn't really a great representation of photogrammetry. Actually - Just noticed Wayne Parris has covered this below a couple of years ago. EDIT : Still a good video. Don't get me wrong.
Great! Would like to see more 3d scanning videos! For Photogrammetry of white objects you can project a colorful pattern with 2 cheap projectors and the result will be much better. It would be great to see if "old" DAVID v4 scanner with line laser (not with SLS projector) is able to scan the chrome skull. High quality industrial monochrome cameras with line laser can solve a lot of the disadvantages from the SLS scanners (e.g. black and shiny).
I stumbled onto this video, I didn't even know 3D scanners were a thing. I'm high and this made me think.. would it be possible that in the future we could have a 3D scanner that could map a certain radius around you in detail? Then you could like walk around the jungle accurately mapping the entire thing, then upload it into a video game with Unreal Engine 10. Could that be a thing??
The "einscan" is nice, but not useable for large parts, without scaling up the whole machine setup. Try fitting a car or building on top of the einscan platform. Or even other unmoveable parts. Both systems have their shortcomings and advantages, and they complete each other. But for most objects, Photogrammetry is WAY more practical.
the idea of photogrammetry is for large scale such as buildings and capturing geological data by using drones. 3d scanner is for reverse engineering like copying products and restoring discontinued parts
Great job Thomas, as ever. I´d like to ask you... which is your opinion about best way (software and hardware) to scan a big volume, maybe, your car or a building? what would you select for that big job?. That is an idea for another future video :-)
Irgendwie siehst du motivierter aus als auf anderen Videos... Tolle Video Qualität freue mich auf dass duet Video (Spoiler: es ist definitiv besser als herkömmliche ! )
I used photo for a LOT of years in Aerospace to measure large tools and we never used it for 3D mapping.
We always used a lot of tie points (fixed points in space larger than the object under inspection that the camera could see from multiple angles. This helps to stitch the photos) We also used scale bars with accurately located targets of a highly accurate distance in multiple locations.
You can use reference points on the scale bars as tie points. The bars need to be in location for the whole photo session in all locations with known good distances.
LOL I almost cried for you when you said it takes roughly 15 min to process the photos. When we first started using Photo, we might have had 200 points to map and used 3 scale bars in X, Y, Z orientation so the software could establish scale in 3D. Add to that a hundred or so tie in points and we could measure 3 stories tall, 300 feet long and 30 feet wide in space to 0.015" (fifteen one THOUSANDTHS of an inch. Or less than half a MM.
Remember this is just points in space, not 3D mapping. We would load the pictures into the computer and start the computing at roughly Noon. THE NEXT MORNING when we came into work the computing MIGHT have been finished! LOL!
Leica came out a few times to try to sell us a scanner setup like what you show but on an industrial size and portable . It never worked for us. It was not as accurate as we needed and we did not need to have a detailed 3D map of the surface so we never did any of that.
We DID use Laser Trackers to do a lot of work and with multiple stations we could map the points in the tool referenced above to plus or minus 0.007" or less than half the error of photo and we could do it in an 8 hour shift or less start to finish.
The laser trackers were easy to use but were not as accurate as Theodolites. Theodolites could measure again to less than half the error of the laser tracker but it took a skilled 3 to 4 man team to do the same job as above.
It is all about the job that needs to be done as to what tool to use. We would map the critical points of joining the wings on an aircraft with Theodolites within the same 7 thou and that is a 3D object 250+ feet wide, 40 feet deep and roughly 4 to 6 feet thick in about 5 hours but again, we were not mapping the contour.
For mapping contour we would use Laser trackers set to measure a point every time the reflector moved a programmed distance, say every one tenth of an inch. We would sweep the reflector over the surface in a grid pattern then take that resulting hundreds of thousands of points and mesh them into a CAD model of what the part SHOULD be in a perfect world.
Uh LOL I guess most of this really does not apply! It is your fault LOL, measurements were my life for over 36 years and I guess I tend to get excited when I see tech being used that I used for so long :D Now that I am retired for 3 years, I miss it and talking about it!!!!!!
If you made it this far.. GOD BLESS YOU! now CARRY on :P
Very interesting. I worked at a sawmill about 10 to 15 years ago, and we used high power infrared scanning lasers (USNR/Perceptron LASAR) to scan logs/cants to determine the "best" cuts to optimize for the most revenue and board footage. The software would build a 3D reconstruction, extrapolate the bottom points, then figure out what boards would fit in the volume of wood based on our daily production parameters. Calibrating those things were a pain-- we also used stepped calibration targets where all the dimensions were exactly known. And even then, we still re-measured them with our calipers to be sure. lol
Too much text, bored after one line, use less word.
This is brilliant stuff! Learned a bunch here and got a confidence boost. 😁
thanks for sharing Wayne, sounds intense!
it's super interesting to read you ! I hope you are having a good time in retirement =)
Baby powder works well if you don't want to ruin your object with paint or anything. Apply on it prior to taking picture or scanning. It does increase the quality on shiny or darkish mat texture.
Very interesting comment man, thanks for sharing.
I’ve had great success with spray on chalk
Lol was just about to suggest that,
Ninja’d
Even better is dry shampoo, it just sprays on and you can get it from the dollar store.
Can anyone explain how to have the final 3D model image of a shiny/ chrome object still have that glossy, shiny but using the chalk spray? I get how you apply it, take photos, create 3D image but do you use post production to add the chrome surface back in after? I’m confused.
Add a polarizing filter to the camera. It will remove most reflections, except for the chrome skull.
Wow, that is a great idea ... I'd like to try that on my Einscan scanner to improve its capability
yeah, you just use a polarizing filter when doing photogrammetry with cars. Also, shoot the photos in raw or HDR mode (a plus for a 'real camera' over your phone) and Lightroom can remove lens distortion for known lenses.
@@JKC40 My Einscan SE scanner has an HDR mode, but I wonder if I would get more complete captures by placing a polarizing filter over each camera ... worth an experiment
Better solution: chalk spray. Works wonders on reflective or clear surfaces.
@@BenVanDenBroeck which brand of chalk spray?? ... I have used baby powder, darn cheap, a little messy to apply, but works incredible well, and is easily washed off. I also have used Krylon "dulling spray", it's cheap, but only helps with highly reflective surfaces, with limited results ... while baby powder helps well with both reflective and dark surfaces.
You are practically one of the few UA-cams I can listen to and believe.
I recommend dry shampoo spray for matting the objects (washes off with water)
Great idea!
You are a real genius 😘
I recommend dry shampoo in general
Be aware of brand and ingredients but it's a classic tip for sure.
thanks nerdy pewdiepie, you helped me a lot bro.
You know.... I woke up this morning not having a clue about this technology and now after your video randomly popping into my suggestions, I really am excited to try this out. Great subject matter!
Tom, I love your channel and thanks for introducing some people to photogrammetry. Those with access to an SLR or at least a tripod with their phone are going to fair a lot better. Keep that Aperature up in the F11 range: the tripod will allow that shutter to be as slow as necessary. Most phones can be set to voice-operation, "cheese, cheese, cheese..." as the tripod gets moved around or the turntable spun with the object.
A light dusting of powder or a polarizer or both helps with the glares. Spray chalk is also great for items that are ok to rinse with water.
Those reading: please join some of the photogrammetry groups online and get ready for a vertical but satisfying learning curve!
Einscan is based on open source technology. You can actually build your own from scratch. Actually, ideal for a Maker's project. The 'parent' project was called DAVID Laser Scaner, and it had two generations - first one used line laser (you had to buy the laser diode at prohibitively large cost of $5), and used motorized setup and video footage to generate vertex cloud, and second generation included a projector shining stripes onto the object with set of photographs - it increased the cost, but removed the need for perfect mechanical setup for moving laser, or at least prism or mirror. Einscan's dual camera setup is improvement over the original idea, btw which sprung around the time Microsoft's Kinect SDK became available. LED projectors can be had now for less than $15 (they have wild resolution of 320x240, but since all it does is display bars, it should be fine), so it's not that expensive.
Sadly DAVID's site was hit with users' data leak at some point, so it had some dark side to it.
Edit: I believe it might have been some time since I looked it up. It had became paid, closed software. What a shame :(
Vatharian do you have any links to the used libraries or resources?
David laserscanner had a free version, but it was not updated for a long time and got more and more limited compared to the pro version, then HP bought the whole David project and rebranded the scanner hardware/software as "HP 3D Scan"
Einscan uses a rotating base with a fixed camera...….I was under the impression that you had to keep the subject still and move the camera around it.
HP purchased David and added it to their private line of products they sell for an absorbent cost. Oops saw already mentioned.
OpenScan I have a copy of DAVID but it’s the free version
It's really sweet of you to thank patreons who support other creators. It's gonna be more important now.
There is sophisticated all-in-one package photogrammetry 3D scan solutions out there as well. Although you don't need it and as you said, you can do it on the go, you always get better results with a calibrated well-known setup. By the way, the technique Einscan is using is called "stuctured-light scanning".
I think if you compared the two techniques of two machines in the same price range you wouldn't notice much of a difference accuracy-wise.
I just have to say that I am a technician at Digi-Key and I spotted that ruler in 0.3 seconds! I was super happy to see that in your video!
For me is photogrammetry all the way, its incredibly versatile cheap and has pretty good precision if you know what you are doing, you can scan from insects to mountains!
I was thinking the same. I was not sure untill i read your comment. Aim a jewelry designer and aim looking for a way to scan small objects. I was thinking about a 360c turntable and take picture with my canon D7. ✌🙏⭐👍💜
Really! Could i know what is the camera and software that you're using? Thanks in advance!
How many scanners you've used? And which ones? Using photogrammetry (for me) is a major nope. Especially if you have more than a few things to scan. My record with Artec Space Spider is about 70 models per workday.
@@ovDarkness I haven't used many scanners in a work environment, and only demoed an artec eva, you know i can't afford a $20k scanner much less $30k for an artec space spyder. Im talking as a hobbyist, i might not get the same precision as one of those professional scanners but as i said i can scan from insects to mountains, underwater, wherever I can take a camera. Hobby grade 3D scanners aren't that good and still very expensive if you want to get any respectable resolution.
Edit: seems the space spider now its under $30k it used to be more expensive, still is a ridiculous price for a clothes iron with 4 modified phone cameras and some LEDs and a mini projector.
@@Vio1007 Sorry im 11 months late YT never alerted me about your comment, i use agisoft photoscan and the camera i use is a canon T3i, when i want to scan very small objects i use a 18-55 lens with a reversing ring, i might add i usually stop down the lens to f11 then remove the lens before reversing it while pressing the show DoF button, that way the lens keeps the aperture while you have it reversed, going to smaller apertures you get more depth of field which is good but you lose a lot of light and some resolution.
I know this is an older video, but I wanted to point out that there are now cellphone apps for photogrammetry that work about as well as shown here, but they generate the model automatically on the cellphone. It's pretty fast, but the detail levels haven't improved much.
Any particular app you'd recommend? For Android, ideally.
Yeah I hate it when people say they are absolutely don’t mention the absolution steer away from all the ones you should still towards
@@lowdrag462 I can't understand, was this a translation? It came out as a sort of word salad.
Why would anyone give this a thumbs down? I think that this video was very objective and told both sides of the story! What else do you want? He doesn't waste a lot of time but gets right to the point.
Hi Tom, good video but i believe the einscan is also "just" a souped-up camera. They can measure the distances thanks to the projected lines and interpolate thanks to photogrammetry too, but this is not how "real" 3d scanners work. Real 3D scanners use Lasers to measure distances and give you a live preview of the appearing mesh and let you know where you have to go over again (yes you have the scanner hand-held and go around the subject to be scanned). Check out the Artec Eva and Artec Spider for "real" 3D scanners
faxxzc this.
Yes, yes, technically, the Einscan is a photogrammetry scanner as well, but structured light is as "real" as it gets. It's actually quite similar to what laser-line scanners do.
The Artec Eva / Spider are also structured light scanners with a camera that works on very similar principles to the Einscan, so i don't know what you find that makes them "real". Yes, they have more in-unit processing and they are faster, but a unit that is at least a magnitude more expensive should be... The Einscan Pro2X Plus has a very similar user experience to the Eva (yes, i have used them both)
Woah I did not at all consider using a drone to fly around a building for photogrammetry. That is awesome!
i brought an einscan a few months back. It works great and turns out 100% usable scans and models. I try the laser scanners and photo ones. They simply didnt work for me.
So i say if you needs a working scanner and have the money the Einscan SE is the way to go.
Also if you want to scan bigger thing its pretty easy to use a tripod and simply rotate the item for each scan.
what was the price
@@2222rehman $1,200 usd
autodesk recap is free for non commercial use and offloads the processing to a cloud
It doesn't look like it's free on the autodesk website?
@@funkycowie definitely is for students and hobbyists. Have a copy myself
Yeah, for non-commercial. Some of the best software is free for non-commercial.
@@MrDead1975 Wait an autodesk photogrammetry program that offloads to the cloud? That use to be called 123d catch.
123d used to be absolutely free if you sign up and you don't care about other people stealing your models, when I tried it out I was disappointed, the quality was way worse than I expected and to get a reasonable model you need to take a lot of pictures.
Anyways, going back to that recap you're talking about, the site literally says you can have a free TRIAL or use it for free if you are a student or educator. But ONLY IF you are one. The licence clearly says: "it can only be used for educational purposes". The terms of service clearly says: "You must be a Student, Faculty, Qualified Educational Institution or Other Authorized Education User (each an “Education User”) to access or use an Education Offering, including any Software licensed under an Education License Type.".
Simple conclusion: it will cost you hundreds of $$$ per year if you're not a student.
You forgot the hobby factor.@@sasjadevries
Meshroom has its own limitations, and you can help deal with smoother surfaces with more photos with a higher res camera.
Good subject! This is quite a question nowadays. I use photogrammetry for anatomical objects and with a 50 Megapixel camera I reach a level of detail that I could never reach with any other method, so for me and for many of us who already own a good camera, i would definitely conclude that for organically shaped objects photogrammetry is the way to go. Not for reverse engineering purposes perhaps and not for real-time capture needs of course.
Which photogrammetry software do you use?
@@@iskyguy007 I use Reality Capture. It is spectacularly faster than others.
@@MakerFarmNLRC is insanely fast compared to the others. But for the hobbyist, the pricing model makes it less appealing. But at work, where I don't pick up the tab, that would be the one to go with. Especially if you have a powerful workstation. What the others would do in 4 hours, I swear RC could do in around 30 minutes. Though I should do another comparison with Agisoft. Supposedly they did a little improvement in speed. And at a one time $180, it isn't as horrible on the pocket book for messing around with at home.
When you say "not for reverse engineering" does that mean you can't get an STL file from your scanning for 3D printing in the end?....surely that is possible with a low resolution iPhone?
Great comparison, i had wondered about the photogrammetry.
I have an Einscan SP and it's so amazing to use for copying things that are no longer manufactured, like brackets or clips.
Thanks for sharing and as always Keep Building👍
would you recommend the EInscan SP? or is there a better 3d scanner for less money in 2020?
@@davidvalens3337 I'm sure There are cheaper better options available now.
Real scanning and phorogrammetry works different.
At work we have a 200.000 euro and 500.000 euro machine for scanning , it is a measuring machine, so it need to be precise.
The object gets a lot of stickers (and coded stickers)on the object, a lot of adapters can be used to capture it in the photogrammetry, this is to capture the middle point of a hole(and a lot more).
The photogrammetry starts, it will not use the pictures for the mesh, it will take pictures of all stickers(coded stickers), and adapters, and the scale bars. Scale bars are for distance measuring(every year these needs to be calibrated 1000 euro). Now you have a pointcloud of only stickers on the object. On the stickers only you can start measuring if you want.
The stickers also good for if camera or objects moves, it will not accept the scan.
Then the scanning starts, the scanner calibrates itself and the scanning is very precise. scan the object, every scan will see the stickers, looks in the pointcloud for sticker, finds it, places the scan on the stickers, this is very precise, after all scans are taken, you make the mesh, the stickers are scant, and are higher, but the software removes the stickers, so the mesh is smooth. Auto cut out, to the cad model, make RPS, and you have the real measurement in real life and very precise, and you never can print this quality.
The camera lenses can be changed 700 mm (capture area)is the big one, and 400 mm is the smaller one, then you get more precise.
We work with blue led light.
Newer versions go to 1000 MM and laser light, even bigger and faster.
You can buy special can of white spray for the objects, it will dissolve after a few ours without cleaning(expensive).
Search on UA-cam for " GOM optical" then you see movies, and how to.
GOM has a free version of their software, where you could use the mesh repair smoothing software, if it is useful I do not know, you can import stl repair and export,
You cannot import a cad model.
Yep, I am a retired tooling inspector and for many many years used photo, laser trackers and theodolites, Industry is nothing like the toys in the video. Two thumbs up :D
Ive been doing nothing but taking 10000s of photo's for models! LOVE THE VID!
Gotta say, the ability to take a photo of just about anything and print it is more attractive then the ability to take a fantastic photo of whatever will fit on the platform
It is indeed sad that EU approved article 13 :( My country is affected too.
Very nice comparison, Thomas! I love how you make your videos!
I used in the past the photogrammetry to scan some household stuff. It is good enough to get a rough model of a thing, but you always have to correct the scan in meshmixer or other software.
Keep up the good work! :) Cheers!
You don't even know what article 13 does, do you?
Never heard of it.
@@squidcaps4308 We are not here to argue about this subject. I know really well what it means. :)
so Einscan here is not a laser scanner but a stereocamera boosting object resolution by projecting reference patterns? that should be applicable into the smartphone scanning game...
Knowing your love for opensource, I'm supprised you didn't include opensource scanner like the cyclop, freelss or my sardauscan (it is hard to calibrate, but it only cost 30$ is much faster than 30min with decent machine and can be used to automate photogametry process)
I've used some laser-line scanners previously and the results weren't usable at all. But I'll make sure to get my hands on one and see how far I can push it!
Some years ago in 1995 I went to a shopping mall in Cape Town, S.A. and they had a laser scanner that took a multi picture map of you while you sat in a chair (very still for 10 minutes) and then it used the info to make or burn a pseudo hologram of your head in a block of clear acrylic about 40mm X 40mm X 50mm. Would this be a similar process to photogrammetry? The image is very clear with very fine detail.@@MadeWithLayers
If a specialized product is NOT better than a manual process the company is NOT doing their job.
Photogrammetry is the best, take hundreds or thousands of pictures to capture some monuments, ruins or castles.
I'm a funboy for that.
Back in a day there was even a developers who worked on the mobile app and the app was creating a 3d point cloud at the moment when you were taking pictures. But I don't know were they got with it, did they finished it or not.
I used agisoft photoscan
Helpful video 👍
This is funny I did my own EinScan vs. photogrammetry comparison. I showed the results in separate videos though, not one real comparison video. I also used a different EinScan model and different photogrammetry software.
Photogrammetry has a leg-up when it comes large subjects, say a large building.
COLMAP doesn't even need to have all the photos be from the same camera or the same day!
Take your own photos, get other photos from on-line sources or stock photos.
Works just fine.
Structured light is a great process but it is limited in terms of object size and access to the object.
Photogrammetry can work with images of a demolished building, assuming you have lots of photos.
Now if only structured light could go back in time.
This is really cool! I've heard of COLMAP! Have you dont this type of work yourself?
Excellent video. Thank you for the information.
4 years later. Cross polarization photogrammetry workflow still the best
Why not stereoscopic light with photogrammetry? Is what forensics use.
Been wanting to find out more about this field for awhile, right now we can make objects but not scan objects, same as the printing press came before the camera, I think this industry will bloom in the near future, thanks for the video Tom!
Industry has been using these processes for over 30 years.. who do you think developed them?
Funny thing is, showing a logo on a box like that Shining 3d logo on a youtube video also applies to article 13 right?
Only the almighty upload filter will know the answer to that question.
No. The amount of confusion around this subject is... embarrassing, really.
The article 13 itself is embarrasing
@@devluz the great filter, another fermi paradox solution :p
Thought about 3d printing ... then thought about scanning and printing something. Einscan SE is like 3.2k at Amazon. Will check back in 5 years!
try adding structured light to the white object (cheap ebay laser projectors that project dots should work)
It that for photogrammetry or the einscan?
@@ryanmickelwait1521 for photogrammetry. the einscan is using a calibrated light source and any light other that from its source will confuse it.
Adding structured light to untextured surfaces is often good enough and always a nicer option than the mess/damage of painting the object.
@@drkastenbrot I guess if you simply use a bath of liquid chalk, then the coating of shiny objects shouldn't be too much of a problem. If you have a chemicaly very stable Objekt, and if the resolution is not of most importance, then you could even try dunkin the object in a CaCO3 solution, let it dry for a fiew minutes, and abuse the fact that the hardening process creates small flakes, and thus creates a perfectly non-smooth surface.
How about a couple of light coatings of hairspray on the shiny surface...…..that would still give you the detail without the shine.@@heron5045
Thomas .. your reviews of the Einscan SE unit convinced me to purchase one .. 4 months later it has already paid for itself through Ebay 3D Printed product sales enabled by this scanner. It does take some time to complete scans of detailed complex parts, but the results are outstanding, and incredibly dimensionally accurate.
What do you sell on ebay?
@@leecaste Top Secret ... have 100% of that market now, not that you want to do anything evil .. but someone else might read this and upend our many hours of hard work. The cool thing about 3D printing is that ANYONE can be a producer, just find a niche that aligns to your own personal interests and knowledge and social media network .. and you will have fun, have free advertising and make a few bucks.
Photo stacking hard to scan objects also works well, but adds an extra set of steps.
Thanks for the great info! If you have Photoshop and probably other photo manipulation software you can also adjust the individual pics removing shadows etc then run it through meshroom hope that helps someone :)
Particularly if shot in RAW
It would be cool to see how well a self build Thingiverse 3D Scanner with you're Smartphone competes against such Scanners like the "Einscan".
I would do that if I knew what to do, but I would use my 15Mp digi camera. Would that be using a video mode to get the multiple photos and at a low resolution to get them all on a memory card?
Skull + dark room = Success ( whit Einscan )
Try again now with the Light Off ( to see if Einscan can do the skull ) thanks
If the skull is dusty or scratched enough, yes. For a pure silver like finish, dark room is not going to help.
Love the colorgrade of your video with with that slight white milky effect. Really smooth and comfortable to watch.
And props for your english. Wizout ze standard görman accent. ^^
2:20 quick idea: spray it with colored plasti dip! you can easily remove it later and it will provide a good rough matte surface to scan
You should take the photos with an actual dslr to get the metadata of the pictures. You can also use photogrammetry with aeral photos like drone footage, which is an advantage. Also the einscan se takes more than 15 min for full color meshes. We have one at my university in trier which i coul use in the past. Still thx thomas for the great vid!
it's nonoptimal to use a phone camera with a rolling shutter for photogrammetry. The phone moves throughout the photo which means geometric contradictions and inaccuracies in the end. A DSLR would be my camera of choice.
Phones also need the OIS turned off, both optical and digital (most can't do this). OIS causes distortions in the image while stabilizing, most people won't notice it, but a computer will. It's most obvious in videos. If left on, many of the images will cause errors in the calculations, and depending on the software, will be left out so you have less photos to do the calculations. Or in others, causes strange results. A cheap point and shoot will do better than the vast majority of smart phones. But like you said, get a DSLR for best results. And if you watch craigslist or ebay, you can get second hand pretty cheap if you keep your eyes open.
The iPhone X and Xs have forward facing 3d scanners. There's some software apps that use the scanner to create a 3D model. Could you do a review of this please?
Maybe you could use a toothbrush to spray paint some black color on the white object. That could give enough texture for the software to find some features.
You can get very low resolution cheap projectors off the internet.
Even some portable ones. Projecting a 'calibration pattern' on a feature and textureless object like the angel should dramatically improve the results, no? After all it is kind of copying the Einscan System.
Yes, that should do the trick. Projecting static random noise (that is, every pixel is at least 50% in lightness and no two pixels have the same color next to each other, no animation of any kind) should be pretty good for the object surface. However, you then need to do another scan for the textures, especially if you use handheld camera. The Einscan system can shot an image with full illumination and continue to next angle.
Great discussion, have tried the hand scanners and a lot of photogrammetry, including drones. I guess I would have to assume that my mobile phone and my drone would be my outdoor choices and a hand-held scanner or other would be my inside choice. also if I had a model that I really wanted scanned nicely I guess I would have to make that choice whether a $1,500 us scanner was worth the scan
If you were to project lines with photogrammetry I imagine that would improve the results (on the white model for instance).
I find that using a good 20 mp camera where I can manually control all the parameters like f stop, shutter speed, exposure and focus gives way better results than a cell phone camera. I put the camera on a tripod and rotate the subject on a simple turntable with controlled lighting. After much experimenting I have gotten good results with photogrammetry. Have been using free version of 3d zephyr software. You can"t beat the price versus a scanner.
Now that is what I wanted to hear...….the simple direct approach with a high res camera.
Photogrammetry includes both processes covered here: metrology through clever photo analysis or through some manner of scanning. In fact, the 3D scanner is processing photographs. Some more appropriate terms might be "structure from motion (sfm) versus structured light," or "multiple view geometry versus structured light."
Also, don't let your output from Meshroom fool you. It may look great, but to brush aside the greater precision and accuracy of the scanner is a mistake. They are not both capturing your subject "just absolutely perfectly." The scanner's accuracy is probably on the order of hundreds of microns, I'm guessing. Your mesh accuracy from Meshlab was probably one or two orders of magnitude worse, and takes usually hours more, to boot, if you use a proper number of photos and have average hardware.
thanks I'm planning on digitizing clay models with a scanner to make a retro fps tribute on steam
Would like to see a review on the HE3D Open Source Ciclop DIY 3D Systems Scanner Kit, something I can afford:-)
I tried photogrammetry and did not have much luck and took huge processing time, but using the Xbox 360 Kinect-1 and skanect, it worked shockingly well. Have you tried Skanect? Trial version was a good, enough for me to buy it (130$ I think).
Very balanced and VERY thorough, quite impressive. Keep up the great work!
"This one's good because it's white, but if it were black we'd have a harder time"
Aand demonetized!
LOL
great :D
13:31
Likely not, because he is not talking about people, but machines and software. If he was indeed talking about a group of people, then without doubt he'd deserve to be demonetized.
@@brianbatie6650 WOOOSH!!!
hi tom ive not tryed anythink like this for a good few years but i did have amazing results in past with davidscanner back then it just used a webcam logitech 920 pro if i remember right and some paper patterns you printed out and put in a corner, it realy did work well i used it at time to scan a model car for a project, might be worth a look how its moved on
how about a circular polarizing filter to cut down on the metallic reflections ?
Bob Thiry then you would have a unicolor surface which is still problematic for photogrammetry
You missed out two categories that would have balanced things somewhat. Scale and flexibility. You can take a camera everywhere, not so much for the Einscan. I have also seen people take scans of large objects where you are limited by size on the Einscan. Still a nice device though.
I collect ancient Greek and Roman coins. Many of us make plaster casts for study and trading by pressing the coin into plasticine clay and filling the mold with liquid plaster. The plaster can be colored if you prefer other than white. This also evens out the color of coins made hard to read due to variations of tone or patina. Have you tried making a plaster cast of a shiny, multicolor or other difficult subject?
In regards to accuracy in photogrammetry, the higher your megapixels the more detail you will capture and in Agisoft for example (around 170$ for the standard version)
you get the calibration tool which eliminate distortion if calibrated properly. you can get sub mm results with photogrammetry
Please do not confuse megapixel or resolution with accuracy. They are almost independent, although if you have freaking resolution you can map on top of image by image with more control points. Still if you don't have something really accurate to relate it to begin, it doesn't matter what you resolve. That is why good photogrammetry tools have excellent scales to begin with. Also accuracy is tottaly dependable on part, so yes you may get sub mm but you can also get sub cm or even sub dm.
Luís Filipe Lopes true, but higher resolution, depending you know what you are doing can also yield better accuracy as the solver have more information to work with. It is true though the higher you go the sharper you need your image to be and here lies the experience of photography in general and photogrammetry as well.
@@-Belshazzar- yes, that is partly correct. First resolution, specially without a fringe light projector, can be misinterpreted, and lead to errors in the first place in terms of mapping, and two points of view, will change their mapping because light doesn't offer sharp information to start from. This is why a pattern, is always better, as it is objective. You can also fake it, it will help. Then, the other part is that, even if you don't have resolution, you can still get very accurate results, in the mean value. There is more associated noise, but quality information is not dependable on resolution per se. Of course, if you add precise scales and targets that can be more dependable no matter the light, than you can achieve a much better photogrammetry model to start from, but in here it lies, it's not dependable of resolution of images, but quality of artifacts. Also, if you don't have enough form (and not texture as it's told in this video incorrectly) the photogrammetry system will loose itself very easy. This is why it's not so difficult to scan small parts that can fit inside the sensor scope, but larger parts are worst.
Dye-penetrant developer may be another solution for shinny stuff. Works good for thermal imaging.
With the skull, coat it in petroleum jelly and roll it in some sand to give it a texture, then the skull can be cleaned up after
I wonder if a phone will ever come out with built in hardware for scanning. Possibly by replacing the camera LED with laser projector and using that to project grid lines during scanning, along with the software to automate the end to end process of building the model.
Just to point out that Meshroom uses CUDA which requires a Nvidia GPU. Not a criticism but just FYI. They have no plans to make it work without one.
@@oudent There are other pieces of software that can do photogrammetry for free that might not need CUDA.
@@oudent Interesting. I am planning on buying a Nvidia card to do this but I may go for a slightly cheaper older gently-used one. I've seen the autodesk one produce good results but also completely fail.
I'm not too worried that he said "probably" but "may have" would have been better and stating what that hardware was would have been best.
Would love to see a comparison between the scanner and one of those "homebrew" xbox kinect.
Photogrametry for me. As an analogy, we can look at the time of VHS vs. Beta? VHS has been more popular since it's more practical and more affordable for the common consumer as well as the industry that caters to the VHS or Betamax crowd. For instance, when one gets into a car accident or need that extra information in a pinch, the photogrammetric images that can be taken in their phones is flexible enough that you don't have to worry about when and where to take out that device just to take all those pictures because your good to go everywhere and anywhere you bring your phone. The 3D scanner, you have to pull out of storage just to use it (you pull it out of storage since it's an expensive device and you don't want just anyone to tinkering with it who don't know how to use it). Going a step further, for crime scene investigation photogrammetry may be more preferable since the investigator taking pictures is already knowledgeable enough with the camera not to warrant an extra step to set up a 3D scanner to mess with any evidence.
I've never tried it but you can try using Polarizing Filters for shiny objects.
Thank you for your intro, the first topic.
What about using a manual turntable or automatic turntable with a cellphone? Would this be a good 3rd option for testing?
12:21 many of the photogrammetry complaints are specific to the software, like Meshroom. For example many of them would not apply to Reality Capture.
Yet another nvidia only software, cuda required...
@@RickBoat That's why I always buy nVidia.
@@3DJapan if you had any thought of using the gpu other than for graphics, sure, but at the time... it seemed like the logical thing to do.
@@RickBoat Ah yeah, I've been using programs that take advantage of Cuda for many years now.
Thank you for the video, very interesting. I have a few comments though. You are not really comparing photogrammetry to 3d scanning. You are really comparing basic photogrammetry to stereoscopic, structured light photogrammetry. True 3D scanners (mostly LIDAR) directly measure the range from surface to sensor (TOF of the laser to points in the surface), rather than indirectly calculating it (which is why the also can directly scale the objects). In general, scale in photogrammetry is arbitrary or very inexact (even when it successfully reconstructs a proper object), because in pure photogrammetry it has no idea how far you were when taking the images, and stereo differences become very unreliable when distance is above a certain ratio of disparity, and projectors really struggle with distance (which makes the patterns unusable).
You sound like you know a thing or two. Could you help point me in the right direction? I want to scan interiors of buildings AND products. I also have a limited budget so am looking for creative solutions. I see people using structure sensor for it, and then you basically have to take the rough point cloud / meshlab and just use it as a template to rebuild the room in a 3d program. This Einscan seems to be more powerful that the structure sensor, but you don't see any talk of people using it for scanning buildings. Something that would provide a more accurate mesh would make workflow quicker, but I would like to keep it no more than 1500 to buy
Personally I prefer laser scanners. You pretty much need to spray with developer either way, but for laser scanners attached to articulating arms, you don't always needs target stickers. Just one reference point if you plan to splice scan data into 1 mesh.
For the shiny skull, dish soap probably works well (many laser users use it) and you just rinse it off after.
Talc powder works great from a sprayer that makes a cloud of dust.
I'm amazed that 3D scanners are still so ridiculously expensive, and limited in size and versatility. How hard is it to attach a couple of reference markers then operate at a much larger scale? Also, did I miss something or did you not talk about texture capture at all?
This depends on the scanner you get. A handheld one could do much larger objects just by walking around them.
@@3DJapanAre there any reasonably priced good ones you know of Phil?
@@Martial-Mat Depends on reasonable. Haha I've seen a few I think that are under $1000 but I don't know the quality level. I've only used the EinScan Pro+.
there are app called Scgann3D , it doesn’t need pc or internet and reconstructs model in 2-4 min (2gb ram/1.5 ghz processor)
The app doesn't exist. Google can't find any information about it.
Wow this was very informative, thanks! I haven’t read all the comments but one limitation I see with einscan is the size of the object. I’d like to scan some r/c airplane models that are up two - three meters in length, I think einscan would have an issue??
ich wusste gar nicht, dass jürgen klopp beim 3d-druck am start ist! XD
danke fürs video.
I love the video Thomas, thank you so much for sharing. I have a quick question. Are there any 3d scanners that can see through hair, fur, or clothing? I know the airports use mm-Wave technology to see through garments, and thermal doesn't really work either. Any ideas?
Really great vid, so useful - as always...
If you can get such good results with a smartphone camera, what kind of results can you get with a high end camera with the right lense, and more pictures? (Yes, I expect it to take a point away from photogrammetry, in the cost category, but I think it'll add one in the performance category.)
Hello Thomas.Why not use VisualSFM for photogrammetry and MeshLab for proccess the 3d meshes? I m doin my thesis at photogrammetry and made a scanner also, very instersting field.
I would be happy to talk to you about photo. I did it in Aerospace for a great number of years. Real life can sometimes be ...... different than what the books say :P
@@wayneparris3439 i ll be happy to hear your experiences.
I can recommend OpenMVS.
It scales well to low thousands of images.
The reduction in actual manual work is something I'm currently working on, because at the scales I'm aiming for manual work is not really an option.
In case anyone here has a camera with an electronic global shutter (fancy video cameras like REDs or black magic and similar usually offer it), I'd be interested in traiding a suitable raw capture for the reconstruction from that capture. Contact me ;)
Does that work for single camera scans? I thought OpenMVS was only for stereoscopic cameras.
First off, amazing video, loved the information in it.
You should do a video showing different techniques to capture hard to capture objects. You mentioned some that wouldn't hurt the integrity of the original model (talcum powder for example), and that would be interesting to see.
One suggestion, that I don't think would affect your rankings or anything, but is important to take into account, is you should have a solid, properly colored background.
I see this all the time in people showing off paint jobs on miniatures, or even just attempting to sell them. If you take the pictures with any organic background, it can affect things like lighting, or just in general affect the image. A solid neutral color background could possibly affect the lighting properties, and allow for a "better" scan. Worth an attempt, or even a "how to" or "do/don't" video for scanning.
What I missed in this test, is how much effort is it to create a watertight model for 3D printing,
The Einscan does it automatic in the software, using Photogramity you would have to do that yourself, and can take some effort to get rid of all unwanted parts, to create a closed model.
Einscan - 5
Photogramity - 2
I use the program Agisoft Metashape for photogrammetric
processing pictures and it can cap off models or fill holes. But this program cost money. (179 dollars) I also use (from work) a 50K real laser scanner from creaform. This is a high resolution scanner that scans up to 2 micrometer depth. But you get penniless only by scanning, it needs special measure stickers costing 150 dollars for 200 stickers. and you need a lot of it. I use +/- 100 stickers for a 2 by 2 meter scanning area.
I watched your other video with the einscan, and it was a pita!
Why in the world would you compare photogrammetry with scanning using a cellphone for the photogrammetry part? That's like comparing bicycles with motorcycles using a single gear bike. Photogrammetry using an actual DSLR (even a $400 mirrorless one) is going to give you results far and beyond what you're getting with that cellphone. This comparison deserves a redo with a proper camera.
I'm not sure this is a fair comparison. Some hastily snapped phone pics isn't really a great representation of photogrammetry. Actually - Just noticed Wayne Parris has covered this below a couple of years ago.
EDIT : Still a good video. Don't get me wrong.
Great! Would like to see more 3d scanning videos! For Photogrammetry of white objects you can project a colorful pattern with 2 cheap projectors and the result will be much better. It would be great to see if "old" DAVID v4 scanner with line laser (not with SLS projector) is able to scan the chrome skull. High quality industrial monochrome cameras with line laser can solve a lot of the disadvantages from the SLS scanners (e.g. black and shiny).
I stumbled onto this video, I didn't even know 3D scanners were a thing. I'm high and this made me think.. would it be possible that in the future we could have a 3D scanner that could map a certain radius around you in detail? Then you could like walk around the jungle accurately mapping the entire thing, then upload it into a video game with Unreal Engine 10. Could that be a thing??
you can find chalk spray to matt shinny objects...
The "einscan" is nice, but not useable for large parts, without scaling up the whole machine setup. Try fitting a car or building on top of the einscan platform.
Or even other unmoveable parts.
Both systems have their shortcomings and advantages, and they complete each other.
But for most objects, Photogrammetry is WAY more practical.
the idea of photogrammetry is for large scale such as buildings and capturing geological data by using drones. 3d scanner is for reverse engineering like copying products and restoring discontinued parts
Great job Thomas, as ever. I´d like to ask you... which is your opinion about best way (software and hardware) to scan a big volume, maybe, your car or a building? what would you select for that big job?. That is an idea for another future video :-)
Irgendwie siehst du motivierter aus als auf anderen Videos... Tolle Video Qualität freue mich auf dass duet Video (Spoiler: es ist definitiv besser als herkömmliche ! )
Einfach mach auf deutsch schreiben hahah