I remember downloading a photogammetry software to the local library computer back in 1998 because the connection was fast enough to not make it take an hour. I then saved it on a 3.5 inch floppy to take it home to install. The example object was a 'shoe' because I guess thats just what you are supposed to do first. I used the digital photo option of a JVC camcorder to create 640 x 480 photos that where then stitched together in a 12 hour process to produce a sub-1000 triangle shoe model, but I was in 8th grade. Looks like the tech has evolved a bit!
I have to say with all the reality capture tutorials I've watched, none have been as informative as yours. "Shoot images, press start hey presto, you got a scan...." I speak for many I'm sure when I say thank you for taking the time to put this together. 🤜🤛
Small tip to go along with shooting on an overcast day; try to shoot some where with a lighter or more neutral floor. The environment will still bounce color around. If you shoot above grass, you'll have a lot of green light bouncing up. A white surface would be ideal, the same light from above will mostly be bounced back up to make the lighting really even.
capturing photos with a dslr in snow covered mountains can lead to far better 3D scan. the snow will light it up like a spherical light dome. You taught it so good! thanks!
Wow. I was involved at a commercial industrial level of metrology. Airbus, Boeing manufacturing etc. I'm astonished at how accessible photogrammetry has become. Great video. Thank you for sharing your expertise ... 🇬🇧
3 роки тому+49
This is awesome, I would be interested in having Blender used on the last part
Wow, I knew you could do 3D scanning but both of those models ended up wild! I got into 3D about 12 years ago and fell out of it and got more into mograph and it's awesome to see where things have come to!
Hey, andrew woow u r 12 years in to 3d, what about 3d artist 3d modelers future are they loose there jobss, if this technology comes.????????.then no future for no 3d modelers?
@@pradeepj7919 They will not loose their jobs. Lots of models are not things that are real that can be scanned. Also these models would more than likely need to be touched up by someone with experience. Also there is much more to 3D than just getting a model into the program so artist are not out of a job.
@@pradeepj7919 Hi, just reading you comment and just wanted to add an answer. I worked in the 3D industry for +10 years too and based on my experience, 3D modellers will not loose their job. This techniques are good for specific purposes (E.g. Historical conservation, education, etc..) but if you want to use this models for real 3D production (E.g. VFx, video games etc..) they will need a lot of job to fix them, like retopologise them, or more advanced: poly flow fixes for animation. So photogrammetry is a great technique but it will not fully replace a full the pure job of building 3D models for specific industries, it will just give you a starting point. Hope it helps!
When I was young I made 3D image with POV 3D programming solution on Atari ST. To see one poor image you to wait 20 mn. With realistic effects it could be hours.
Thanks for this easy-to-follow tutorial, I've actually struggled with this in the past so great to actually see it play out step by step. How do you like the new iPhone lidar scan stuff do you think it holds up to your method here?
Thanks! The LIDAR is promising, and I'm excited to see where it is in 3-4 years. As is, it only shoots out a few hundred dots per burst, and they're spaced out several inches at typical scanning distance. With photos, you can get ~12 million points of info per photo, ie sub-MM spacing between info points. LIDAR definitely works better for shiny objects, which is why I'm excited for it, but if i were to do this shoe with Apple LIDAR it would basically be a blob. Fingers crossed for the future
@@patrickletourneau8177 Good to know, I'm an android guy myself but was waiting on the next iPad Pro so I could mess around with the lidar stuff but seems like it's not really there yet huh. I'll try this method as well. Thanks for the info 🤙🏾
F-n amazing! Get this tech to the police department for use in crime scene forensics. Detectives could put on virtual headsets and see the crime scene anytime they needed. You could bring in jurors to see the crime scene in the same way!
Hello good people, i need to scan an engine for my project and i wonder how dimensionally accurate the 3D model from this method will be. Is there a way to specific dimensions to make sure that all sizes are correct? Thank you in advance!
You said it would measure the absolute depth of each pixel. It seems that his isn't correct and it's actually relative depth (computed from parallax angles) with some kind of approximation of scale based on the TOF range sensor in some cameras (if the camera includes that info as metadata in the image file). Not all cameras have that. And some cameras have a very poor resolution or accuracy that cannot be relied upon to actually measure an object. So what I think you get is at best a poor approximation of absolute depth. Put it another way, let's say I have a bracket I need to reproduce on a 3D printer. The bracket includes angles and things like screw holes that need 0.5 mm resolution and 1mm accuracy in all surfaces. I'm guessing that the mesh that is created will not be dimensionally correct in all axes. If I'm wrong, please let me know.
You might consider using polarizers with ring-light photos. Cover the ring light with a polarizer, and orient another polarizer on the lens at 90 degrees to cut most of your specular reflections and reduce the required exposure latitude overall!
With Nanite in Unreal Engine 5, we may be able to use these meshes without (too much) cleaning (And Reality Capture got acquired by Epic, coincidence?)
@@chrisrossi5176 I’m using a dual Xeon windows PC. I’m not sure if meshroom is offered for Mac. All you need to do is google meshroom and I’m sure you’ll find it. I will say you need an Nvidia graphics card for meshroom. It’s free and open source.
I haven't seen the whole video but I'll know it will be interesting I'll just subscribe and save the video for later. Thank you for your contribution and knowledge and thank UA-cam for creating this platform.
In my experience, Meshroom has a much harder time especially with turntable workflows. RealityCapture and Agisoft are the best choices for professional output. Perhaps next time we can cover Meshroom, though!
I wish there was a way you can just scan the shoe with the camera and it captures everything including the color of the shoe and then you can go back and do trimming and editing.
Imagine taking images from all angles from your (past) girlfriends and creating 3D models out of them... Great showcase on how this works! 3D modeling, and working with 3D programs like Blender is such an awesome thing that I wish I had more time to learn it.
I want to use a drone and multiple cameras to made 3d inversion type maps for live-ins, handicapped or just people who don't have time, to see parks, streams, likes and nature in general. I'm a retired surveyor(40+ years), and I understand photogrammetry from that standpoint. I don't want to make money(I think), I just want to make some people's dreary days, a little more fun to get through ...
Great video! Question: Do you do something special with the RC and turntable process? When I stay static, taking photos from same position and only my object turn on, i get problems with my align. Thank you!
What Android App is being used? I searched the comments and found at least a dozen people asking the same question. For now, I am using Open Camera. Let's see how well it works out.
Iconasys makes very nice photography turntables that can take up to 8,000 photos per rotation if necessary. The automated photography software and background removal tools make the image capturing much easier.
Are there ways to attach small adhesive pieces of glossy paper or other reflective material or perhaps designated special purpose markers to assist in 3D image reconstruction? I've seen people use such markers/stickers with a 3d scan to produce high precision models.
I believe that would be more useful if you have a surface with a flat color and no anchor point for the align process to rely on. For instance if you want to scan a room with a white wall it's gonna be difficult to get the correct position without putting marks on it
I reckon that would work if you can control the lighting. With outdoor light you inherently have shadows. So turning the object around would cast shadows differently at every angle. Not sure how the software would handle that. Would be cool if the software could recognize the difference in shadows and cancel the unwanted lighting information.
Gone through this process twice now and each time my box is always empty after the alignment is finished. So where you see your shoe I don't see anything except the box, grid, the pink green and blue lines in the middle and some of the little cameras. But no model car in the middle. What am I doing wrong?
Halfway through and this looks awesome for reverse engineering. If you had say a 5 sided enclosure where the open end is the bottom and the enclosure has features inside like two screw posts and a few indentations, can you capture those features by going all the way under it and capturing it? Also can this export an stl file that can be converted to a dimensioned model in Fusion360 or Solidworks where you could edit out whatever the object was mounted to for the setup?
Iconasys makes very nice photography turntables that can take up to 8,000 photos per rotation if necessary. The automated photography software and background removal tools make the image capturing much easier.
Is there none CUDA based freeware software available for the few times I want to play with this just for the sake of it? Can't find anything with acceptable results
I remember downloading a photogammetry software to the local library computer back in 1998 because the connection was fast enough to not make it take an hour. I then saved it on a 3.5 inch floppy to take it home to install. The example object was a 'shoe' because I guess thats just what you are supposed to do first. I used the digital photo option of a JVC camcorder to create 640 x 480 photos that where then stitched together in a 12 hour process to produce a sub-1000 triangle shoe model, but I was in 8th grade. Looks like the tech has evolved a bit!
Wow, I had no idea PG was around in the age of floppies! Thanks for sharing
The youtube algorithm finally gave me something i needed, amazing tutorial 👍
😂😂👍🤘
Oh man I’ve always wanted to be blessed by The Algorithm. Glad you enjoyed it!
Yes!
Good for you, Creator.
Same bro
I have to say with all the reality capture tutorials I've watched, none have been as informative as yours.
"Shoot images, press start hey presto, you got a scan...."
I speak for many I'm sure when I say thank you for taking the time to put this together. 🤜🤛
Small tip to go along with shooting on an overcast day; try to shoot some where with a lighter or more neutral floor. The environment will still bounce color around. If you shoot above grass, you'll have a lot of green light bouncing up. A white surface would be ideal, the same light from above will mostly be bounced back up to make the lighting really even.
capturing photos with a dslr in snow covered mountains can lead to far better 3D scan. the snow will light it up like a spherical light dome. You taught it so good! thanks!
Wow. I was involved at a commercial industrial level of metrology. Airbus, Boeing manufacturing etc. I'm astonished at how accessible photogrammetry has become. Great video. Thank you for sharing your expertise ... 🇬🇧
This is awesome, I would be interested in having Blender used on the last part
Yeah just make a model from 6 pictures.
Perhaps #TomDickson could post a vid on this.
With Meshroom + Blender this can all be done for free.
Wow, I knew you could do 3D scanning but both of those models ended up wild! I got into 3D about 12 years ago and fell out of it and got more into mograph and it's awesome to see where things have come to!
Hey, andrew woow u r 12 years in to 3d, what about 3d artist 3d modelers future are they loose there jobss, if this technology comes.????????.then no future for no 3d modelers?
@@pradeepj7919 They will not loose their jobs. Lots of models are not things that are real that can be scanned. Also these models would more than likely need to be touched up by someone with experience. Also there is much more to 3D than just getting a model into the program so artist are not out of a job.
@@pradeepj7919 Hi, just reading you comment and just wanted to add an answer. I worked in the 3D industry for +10 years too and based on my experience, 3D modellers will not loose their job. This techniques are good for specific purposes (E.g. Historical conservation, education, etc..) but if you want to use this models for real 3D production (E.g. VFx, video games etc..) they will need a lot of job to fix them, like retopologise them, or more advanced: poly flow fixes for animation. So photogrammetry is a great technique but it will not fully replace a full the pure job of building 3D models for specific industries, it will just give you a starting point.
Hope it helps!
That old shoe got a lot more exposure than was ever expected, I'd say! Thanks for the content :)
90% of the time I have no idea what you are talking about but I watched till the end and subscribed as this is totally mesmerising 👍🏻
Which mobile app do you recommend for photos?
What was the name of the camera/photo app you used in the first couple minutes?
This tutorial is so well done. More from Patrick, please!
This has blown my mind.. I didn't know how easy and accessible this was.
If accessible means paying 3750 euros....
@@barderino5673 3750 Euros for what?
I totally appreciate this tutorial but if its possible to make one through blender would be amazing!!!
Yes Blender please!!
You can clean up things with blender, but you cant "scan" here
This is awesome! Thanks for putting this tutorial together!
What app did u use to capture the photos?
When I was young I made 3D image with POV 3D programming solution on Atari ST. To see one poor image you to wait 20 mn. With realistic effects it could be hours.
I do it too and I made a 2D scan on printer lol now we can scan 3D asset with a phone I love technology 🤣
Thanks for this easy-to-follow tutorial, I've actually struggled with this in the past so great to actually see it play out step by step. How do you like the new iPhone lidar scan stuff do you think it holds up to your method here?
Thanks! The LIDAR is promising, and I'm excited to see where it is in 3-4 years. As is, it only shoots out a few hundred dots per burst, and they're spaced out several inches at typical scanning distance. With photos, you can get ~12 million points of info per photo, ie sub-MM spacing between info points.
LIDAR definitely works better for shiny objects, which is why I'm excited for it, but if i were to do this shoe with Apple LIDAR it would basically be a blob. Fingers crossed for the future
@@patrickletourneau8177 Good to know, I'm an android guy myself but was waiting on the next iPad Pro so I could mess around with the lidar stuff but seems like it's not really there yet huh. I'll try this method as well. Thanks for the info 🤙🏾
@@JonathanWinbush Lidar is good for room-size scans, but not object size scans, if that makes sense.
@@thegreatszalam ya 100% thanks
I'd expect the lidar to be used like the AR camera for hitfilm. You can track your 3d environment and place 3d models in videos.
F-n amazing! Get this tech to the police department for use in crime scene forensics. Detectives could put on virtual headsets and see the crime scene anytime they needed. You could bring in jurors to see the crime scene in the same way!
Law enforcement already do this with LIDAR scanners, super neat!
Then we get minority report...no thank you haha
incredible video esp with the commercial comparison at the end!
Thanks Patrick for the video, didn't know about this method, much appreciated the sharing. Keep rocking with more tutorials like this.💪
Cloudy day = worlds biggest soft box.
It's amazing how far we can go with just a smartphone.
Hello good people, i need to scan an engine for my project and i wonder how dimensionally accurate the 3D model from this method will be. Is there a way to specific dimensions to make sure that all sizes are correct? Thank you in advance!
that's absolutely fantastic, can't wait to have a go and try it. can you do some more samples?
is there a good software alternative for photogrametry that is free? thanks, great tutorial!
try meshroom
3DF Zephyr
Meshroom
Meshroom is free and open source; I had a bit of difficulty compiling it myself because of nvcc, but once you get it set up it works like a charm
3dim capture you do it straight on your phone and it’s free but the scanning is not the best
You said it would measure the absolute depth of each pixel. It seems that his isn't correct and it's actually relative depth (computed from parallax angles) with some kind of approximation of scale based on the TOF range sensor in some cameras (if the camera includes that info as metadata in the image file). Not all cameras have that. And some cameras have a very poor resolution or accuracy that cannot be relied upon to actually measure an object. So what I think you get is at best a poor approximation of absolute depth. Put it another way, let's say I have a bracket I need to reproduce on a 3D printer. The bracket includes angles and things like screw holes that need 0.5 mm resolution and 1mm accuracy in all surfaces. I'm guessing that the mesh that is created will not be dimensionally correct in all axes. If I'm wrong, please let me know.
You might consider using polarizers with ring-light photos. Cover the ring light with a polarizer, and orient another polarizer on the lens at 90 degrees to cut most of your specular reflections and reduce the required exposure latitude overall!
Yeah i shoot with an ar400 right light and a custom polarizer from the master, Erik Christensen
With Nanite in Unreal Engine 5, we may be able to use these meshes without (too much) cleaning
(And Reality Capture got acquired by Epic, coincidence?)
I didn't understand a work of it, but enjoyed watching.
imagine being his neighbour and seeing him make 100 photos of a shoe in his backyard
Smart people are always seen as crazy by idiots
"Come and look, love. He's trying to sell his single smelly gym shoe on OnlyFans."
@@BeachBobRoss lol
Any plans for a tutorial on how to make that pro scanning setup?
3:38 so it's not gonna work well with AMD cards will it? What's a different program I can use then. Preferably free
I don't know about supported card types but 3DF Zephyr or Agisoft Metashape are good alternatives.
Metashape is good with AMD
@@rskl3284 is it free tho, like meshroom free
I really like your delivery.
I’ve used meshroom with my iPhone to good results.
As long as there’s features your fine.🤓🤜🏼🤛🏼🇦🇺🍀
Did you find meshroom on the apple app store at all pls? Doesn't find it on mine :(
@@chrisrossi5176 I used my phone to take the photos and used meshroom on my workstation.
@@stevesloan6775 Thanks for the reply Steve, where did your grab meshroom from pls, was it for pc or mac?
@@chrisrossi5176 I’m using a dual Xeon windows PC. I’m not sure if meshroom is offered for Mac. All you need to do is google meshroom and I’m sure you’ll find it. I will say you need an Nvidia graphics card for meshroom. It’s free and open source.
@@stevesloan6775 Thanks for the info Steve, will have a look for it, much appreciated.
Very interesting. Any idea how well it would work using stills from a drone? I’d love to create a model of my property to help plan my landscaping.
It works great, and I’ve done exactly that!
I haven't seen the whole video but I'll know it will be interesting I'll just subscribe and save the video for later. Thank you for your contribution and knowledge and thank UA-cam for creating this platform.
ok thanks, that's a pretty great comparison
I'm not sure when I am going to use this but I'm glad to have learned something great
How about meshroom? Open source software, no payments needed
Yeah 👍🏻
In my experience, Meshroom has a much harder time especially with turntable workflows. RealityCapture and Agisoft are the best choices for professional output. Perhaps next time we can cover Meshroom, though!
@@patrickletourneau8177 I'd like that a lot
man what a great time to be alive
I wish there was a way you can just scan the shoe with the camera and it captures everything including the color of the shoe and then you can go back and do trimming and editing.
Its awsome. I can predict this is the basic technic to build communication via hologram
Microsoft Mesh was announced a few days ago
the description is missing!!!!!
Imagine taking images from all angles from your (past) girlfriends and creating 3D models out of them... Great showcase on how this works! 3D modeling, and working with 3D programs like Blender is such an awesome thing that I wish I had more time to learn it.
VERY cool video. Blew my mind...
How did you capture the shoe bottom in your turn table setups? Does the shoe laces keep their positions if you till the shoe?
I want to use a drone and multiple cameras to made 3d inversion type maps for live-ins, handicapped or just people who don't have time, to see parks, streams, likes and nature in general. I'm a retired surveyor(40+ years), and I understand photogrammetry from that standpoint.
I don't want to make money(I think), I just want to make some people's dreary days, a little more fun to get through ...
Nothing wrong with charging a fee for something valuable that took time and energy to create. That’s how the world has worked for thousands of years.
holy shit! the shoe looked perfect
Great video!
Question: Do you do something special with the RC and turntable process? When I stay static, taking photos from same position and only my object turn on, i get problems with my align.
Thank you!
Are you removing the background from the images, only leaving the product or asset behind?
Can you do this in Blender 2.92 ?
Did you find out?
@@Max-bi9nf no
Wicked!!! Secret crime fighter!? Vigilantes unite!!!(in 3D nerd land 😆🤣)
This is awesome - thanks for sharing this!!!
Thanks Patrick! You rock!
Ahh.
Expensive subscription software. Yeeepiiiieee
Wow! Did an alien invent this software? Amazing!
The HUAWEI mate 20 pro can do it without any program. But this is awesome. Thanks
Fascinating stuff, and I'm not even into photogrammetry or 3D modeling or anything like that.
Can your 3D program of choice be Element 3D?
Thanks a lot, this is a great tutorial! Learned a lot.
is it possible to open a reality capture finished 3d model without installing any programs on your android phone ???
the future, almost startrek holograms, lets do it>
Omg, that was amazing when it render in2 a sneaker w/ color. Do U have an online tutorial on this app? Thx 4 sharing on Utube.
What Android App is being used?
I searched the comments and found at least a dozen people asking the same question.
For now, I am using Open Camera. Let's see how well it works out.
Wow, what an introduction! Great tutorial thank you!
really nice thanks!!
Glad you enjoyed!
You’d get better coverage by keeping the camera fixed and rotating the object as far as that is po.
Yep but turntable workflows are a bit much for an introductory video, I thought
And I love the future 😍😍😍
What Android photography app are we using?
Ohhh man... I cant wait to try this.
Nice to see you working
Educational. Thanks for sharing.
What is "pro app" he mentioned to make photos?
01:30 Ah, the infamous "shoe phone". Getting smart!
This is insane!! So cool!!
Clear and understandable thank you so much!
advice on which camera to use?
why not to take dslr shootout in overcast outdoors, or set proper soft light in the studio?
Very detail video. Thanks
Great tut! If someone wanted to offer this as a service what do you think one can charge a client to do a shoe like the one demoed? Thanks
Looks amazing. Can you do a video on bringing this into unreal engine 5 and perhaps using photogrammetry with nanite?
can you post a gear list for the turn table setup?
Iconasys makes very nice photography turntables that can take up to 8,000 photos per rotation if necessary. The automated photography software and background removal tools make the image capturing much easier.
Are there ways to attach small adhesive pieces of glossy paper or other reflective material or perhaps designated special purpose markers to assist in 3D image reconstruction? I've seen people use such markers/stickers with a 3d scan to produce high precision models.
I believe that would be more useful if you have a surface with a flat color and no anchor point for the align process to rely on.
For instance if you want to scan a room with a white wall it's gonna be difficult to get the correct position without putting marks on it
amazing tutorial! in the normals part of the tutorial, I can't find the link to 'reproject model texture' help!
Wow! Really great! This is so accessible, love it
So what would happened when t subject is on a black stand w/ black background. Would it still captures the toilet stand?
Can this be transfer onto 3d printing printing machine for printing?
Sure can, I've printed a number of scans
@@patrickletourneau8177 Hi can you name the model of your 3d scanner used in this video?
Would it be less work with the object on a monochrome surfaced lazy-susan, and the camera on the tripod?
Yeah with a black felt background and lighting directly behind the camera. Check out `Scanning in The Void` on UA-cam for that workflow
I reckon that would work if you can control the lighting. With outdoor light you inherently have shadows. So turning the object around would cast shadows differently at every angle. Not sure how the software would handle that.
Would be cool if the software could recognize the difference in shadows and cancel the unwanted lighting information.
@@boetalouz Agisoft De-Lighter is exactly made for this
Me that don't know anything, what would be the purpose of a 3d vs a 360view video of the same object???
Gone through this process twice now and each time my box is always empty after the alignment is finished. So where you see your shoe I don't see anything except the box, grid, the pink green and blue lines in the middle and some of the little cameras. But no model car in the middle. What am I doing wrong?
Just wow !
That was amazing!
Halfway through and this looks awesome for reverse engineering.
If you had say a 5 sided enclosure where the open end is the bottom and the enclosure has features inside like two screw posts and a few indentations, can you capture those features by going all the way under it and capturing it? Also can this export an stl file that can be converted to a dimensioned model in Fusion360 or Solidworks where you could edit out whatever the object was mounted to for the setup?
Iconasys makes very nice photography turntables that can take up to 8,000 photos per rotation if necessary. The automated photography software and background removal tools make the image capturing much easier.
Great tutorial!
What is the lowest size possible of the output file enabling zippy loading of web pages?
Is there none CUDA based freeware software available for the few times I want to play with this just for the sake of it? Can't find anything with acceptable results
Does this require an aftermarket graphics card and does it accept Canon RAW images?
Super kool magnific👍👌💎💎💎💎💎
Great video, (as always). I started posting my soft on youtube because of you, but if sobody could give so feedback then i know
Hi, thanks for sharing. Can you get accurate dimensions as in 3D scan? Would it be reliable for reverse engineering?
I want to know this as well.
Yep
Thank you so much!