My grandfather was a photogrammetrist back in the 60's until he retired in the late 80's, He died when I was young but it's amazing to think that he spent half his life working with the same mathematic formulas they use for measuring photos to print models in this video lol
Bookmarks for me. 0:30 Recap 02:10 Taking Photos 04:15 Standard Reconsturction 04:50 Hit Start Or Structur for motion out Tip 05:30 Troubleshoot capture issues 05:45 when Happy Complete Cpmputation 05:55 Notes Turn Green as they c0mplete May take a while 06:12 Double click TEXTURE BUTTON 06:30 Export Model to Edit 06:45 Augmented Reconstruction 08:14 Live Restruction 09:10 All Mesh are Hollow-Plane Cut 09:30 Change Scale 09:43 Sculpting 10:03 Wiki Wiki Time! step 1 Drag photos into Programme Step 2 Right Click Structure from motion-Compute Step 3 Wait until computed Step 4 When Happy Finish Compution Step 5 Double Click Texturing Note to View the model Step 6 Export Model Step 7 Plane Cut in 3D sotware Step 8 Change Scale to suit
You prolly dont care at all but does anybody know of a trick to get back into an instagram account? I stupidly forgot the login password. I appreciate any assistance you can give me.
This is the most amazing thing I've seen in at least 24 days. Now I don't know if I want to 3d print something or put it in Unity... or whether I want to mod my house into GTA5... or build a rig that automatically rotates the camera around an object.
@@talharoxen you mean putting the 3d model into GTA? You use the mod tools, I forget which one you use to export to rockstar format but it should be easy to find. There's 5 or so tools in total that help you mod, it's not as easy as a Bethesda game but it's not that hard thanks to lots of documentation and community support.
Note: Always take note of what focal length you're shooting at - take not of what lens, focal length, camera type and settings you're using. If your scan comes out wrong it can help you when having to enter metadata manually if a problem occurs. It's especially important to make a note if you're using a smart phone camera because often vital data such as focal length is not recorded in the metadata - which is some cases makes the pictures unusable if you can't figure out the focal length and enter it manually. I can strongly recommend people using the following trick: Import the photos into a program like Photoshop or GIMP and add a mask around your object removing detail in the background. You can also zoom in if you couldn't get close enough to the object in the first place. Remember to retain all metadata while exporting your photos.
I did this with a local cathedral (with permission, of course). My results then weren't great however, so I may have to try reprocessing the images with this software.
There is a Litchi software for DJI drones that can do that. You just set the waypoints and camera angle and it should work (but be careful if you use that)
@@logitech4873 you either need a better camera or fly closer. If you used Photoscan or even RealityCapture, you will not see improvements in other software.
It has advantages, but also disadvantages. This is why we combine different scanning techniques for perfect results. Great video and great progress in the software!
@@mageenderman Well, I dont work with photogrammetry, but I have read that they also use laser scanning. Its the most precise way to scan a object and the results are outstanding. Here is a video by Konami scanning Schalke 04 stadium and showing previews from the game: ua-cam.com/video/qAkUF4Kbgxk/v-deo.html
you might be interested in GL Intercept github.com/dtrebilco/glintercept here is one possible application (note this guide is quite outdated): brian-horning.com/google-earth-to-obj-using-glintercept/
He touches on it in the video, but the single most important thing you can do to achieve high-quality results is to use a good camera. Yes, you can get good results with a Gopro, cellphone or point and shoot cameras, but if you compare a model generated from a GoPro to a model from a DSLR/mirrorless camera using a good prime lens, the results from DSLR are absolutely superior. When comparing models between lower-end cameras and high-end cameras, look at sharp corners/edges, photogrammetric reconstruction with lower-end cameras struggle to produce sharp corners, they come out more round. If possible, use a full-frame camera. Make sure you move between shots, no standing and pivoting. Take steps in between photos. Take the photos in a systematic way: go low, medium, and high all the way around. For tall structures, you can also scan vertically (take photos as you move up the structure, scoot over a bit then come back down until the whole thing is covered). You want to disable autofocus as it causes focus changes between photos and will cause pixel blur when the algorithm looks at different photos of the same feature. Set focus to infinity and leave it there. You want to achieve high overlap between successive pictures, both vertically and horizontally (like 70% to 80% overlap). More overlap means more matched features means a more dense point cloud which means a tighter mesh. If you want to know how good a job the software is doing with your photos, turn the textures off and look at the raw surface. The texture hides lots of artifacts. Use a fast shutter speed as well to take motion blur out of shots if your shooting by hand and not with a tripod (this is especially important if shooting from a drone). Use a polarizing filter to help reduce glare if shooting outdoors. Like hard shadows, the software struggles with how to handle it. Post-process your photos to increase contrast in shadowed areas.
If it's a small object you could use armature wire to hold it up so that you could see under it. In any case, try to put the model on a base that is smaller than the model.
Awesome, i know exactly where that first statue is, one of my favourite spots in Prague. I go to the cafe in the National gallery and that restaurant to the right all the time :)
Awesome video! Would be nice if you do a live tutorial on 3d reconstruction of more problematic objects, like ones with shiny and flat surfaces without texture. I.e. like reverse engineering old injection molded parts with glossy surface and flat color, or something like this.
Just when my modelling abilities were starting to improve, this comes along and makes modelling obsolete. I've wasted my life. It is rather cool though. Can't wait to try it!
Hey they're not obsolete! For mechanical parts, CAD is still the best option. And for organic shapes, you'll be really happy to know basic modeling when working with photogrammetry meshes, they almost always need a bit of editing/clean up before printing :)
@@Prusa3D It's good to know I'm not worthless just yet lol. It's funny you should mention CAD, as I recently started using CAD to design machines, and have also sucessfully made game assets from CAD models. All of the metal targets in my FPS game are actually real-world products that I produce. The game models are just slightly reworked CAD models.
Thank you for the video. I was new to the topic and tried to read up on it, but your video was just perfect to get into it. Especially the hint that you can make ideal photos on cloudy days has helped - on cloudy days it is not lacking here with me anyway...
Great video!! Thanks for introducing Meshroom to us and a tutorial on how to use it for photogrammetry :) I was going to start doing this soon with Autodesk ReCap, but this might be a good alternative as well.
Oh wow, I've recognized the accent and the statue of Dvorak, then I found the video about your printing farm. Great job you're doing there! I wish I could visit the farm one day and film interview with you!
Yes meshroom is great! I tried to use meshroom at work to create a reference model of a complex assembly, but management was too stupid to understand why it was necessary. Let's just say the original documentation was useless for building something with thousands of parts. A 3D model of the entire assembly could've been built with photogrammetry in a couple of days, then distributed to every assembler for reference. NASA did something similar for their space craft where they always built two of everything. Photogrammetry is a much more affordable alternative to this.
@@dudeabides23 - Indeed. I learnt photogrammetry from a guy called Sid Ray, who I suspect had a fair bit to do with military photographic interpretation around that time.
There's still plenty of room for improvement, especially when it comes to speed and ease of use. But yes, it's pretty incredible how fast did the technology improve :) - Mikolas
Very cool. thank you for letting us know about this program. I've not been super happy with photogrammetry since 123d Catch was taken down and this looks like it may get me back in to it again!
The live reconstruction might be fun if you have some sort of file syncing tool. Take pictures on your phone and have the model already processing by the time you get home.
The traditional way of 3D modeling real-world objects is time consuming, so an automated, easy and affordable way of using photogrammetry that reduces production times where 3D modeled objects are used is a good thing.
Imagine how far we will have come in 20 years. Conveniently scanning everything within seconds with our "phones", if they are still called like that in the future.
Wow, uprime videl sem thumb asi pred 3 tydnama nebo kdy a az ted sem se k tomu zase nahodne dostal a koukam ono to ma pres mega sheldnuti a je to cech :D ... no nic, tohle pokud se mi vysledek podari dostat az do UE4 tak to bude super. Diky moc za info ohledne meshroom!
Hello all, if you are trying to do a really precise close-range 3d scan using photogrammetry, cover gloss, transparent, and homogeneously colored surfaces with texture, such as random lines drawn over a masking tape or a checkered cloth with fine details. Otherwise, the software cannot determine the depth and small details. They do ok but you cannot chase millimeter level detail. Also, when you import .obj files to a "CAD" software, you won't have a texture to help you, just a mesh comes with its faults. If you plan to do some stuff on the mesh, place some marker bumps or lines on the surface to help and guide you. If you know more, you are more than welcome to contribute to this post. Thanks!
Some advice on photogrammetry: 1-Use softboxes to illuminate your scene...lights must be flat and casting the least hard shadows as possible. 2- use a solid background..white, black or green..ive gotten good results with blackand very little image retouch. 3- mask or use chroma key on your pictures..this will eliminate all the innecesary information from the image, letting software create tie points only from the points of interest. 4- use a DSLR..really. Most DSLR equipments include important exif data such as focal length, iso, shutter speed and the type of lens used...some of them can even estimate yaw, pitch and roll..which helps most photogrammetry softwares compensate errors in the alignment procedure by using such camera info...effectively tying more points from the pictures overall. 5- use a turntable...keeping your camera fixed in one place and rotating the object of interest can produce amazing results..but due take in account you must forcibly mask your picture...this process can take some time...take in account that there is a point in which taking more pictures will not make a significant difference on the fidelity of your model..you dont need to go up to the fivehundreds or thousands of pictures for something you can easily reconstruct with only 30 to 50 pics. Keep it simple. 6- Postprocess your photos...i recomend using adobe bridge with the camera raw add-on to increase the contrast of multiple pictures simultaneously..this will help software distinguish borders more easily and an overall result in a less noisy point cloud. Just remember to not clip your highlights or crush your shadows too much. 7- Reconstruct in a low or medium quality so you can save time...if you are conviced with the results..simply repeat the process in a higher setting..this will help you save extra time..since medium reconstructions can take only minutes, you can realize faster if the photos you shot were good enough to really reconstruct your model or if they were a complete miss.. 8- Play with your settings...overall experiment...photogrammetry depends on multiple factors...photo rig..lighting..the complexity and material of the model..background..image noise..photography postprocessing...keep tweaking until you amount enough experience to realize how to have smoother workflows. 9- You will make mistakes, learn from them...my first reconstructions looked like chronenberg monsters, very little tie point clouds, bad lighting, bad photo quality...you will learn from these mistakes along the way. Today I can reconstruct models up to a million+ poly count with realistic textures and so can you. The important lesson here is that you learn a long the way and keep trying new things.
Unfortunately, we can't agree with some of those. Especially for Meshroom! 2,3) Do not use background and definitely do not chroma key your pictures, this decreases the quality significantly. We talked about it with the developers: "the results are worse because we will only have points on the object which represent a small sub-part of the image, so the camera parameters are less constrainted which leads to less precise results" 4) EXIF data are included in pictures from phone for almost a decade, we got excellent results just by using smartphone camera 5) No turntable with Meshroom
I love doing photogrammetry. The only thing that I wish was different is that there were more software that runs natively on a Mac. I know Macs are a smaller part of the overall market. Before anyone says "Get a Windows machine!" I'm planning on adding an Nvidia card to my Mac Pro and make my hard drive dual boot to have both macOS and Windows. I just know Mac better than Windows right now.
For those that dont have Nvidia & don't have to means to get a graphics card. you can use a virtual machine in Microsoft Azure that has it. it comes about 60cents per hour
I have an idea, with the same setup with the white background, just use a motorized rotating plate in the bottom, and use an automatic shoot every 1 sec
I've tried Meshroom because I wanted to do a tutorial on it but I found it complicated and really had trouble getting good results. These days I use Reality Capture. It's not free but pretty cheap.
I've used PG on old footage from movies or documentaries... Comparing the models made from old footage with models made of the same scene with newer footage reveals subtle changes in shape. Besides buildings and construction, it is fun to see a hillside 'heave' during the rainy season and 'shrink' by the end of summer. Sure, it's only a few feet out of thousands of cubic yards, but is can be measured! Similar things are seen with Grand Canyon footage; old fly overs and newer ones show lots on interesting 'shifts' in form. Are volcanologists using this?
Is there any way to use this technology with an older mac? i have a MacBook Pro (13-inch, 2017, Two Thunderbolt 3 ports) All Photogrammetry programms ive tried wont work with my system. Very grateful for any tips, or programms that i could use
My grandfather was a photogrammetrist back in the 60's until he retired in the late 80's, He died when I was young but it's amazing to think that he spent half his life working with the same mathematic formulas they use for measuring photos to print models in this video lol
Do you have some examples of his work?
Bookmarks for me.
0:30 Recap
02:10 Taking Photos
04:15 Standard Reconsturction
04:50 Hit Start Or Structur for motion out Tip
05:30 Troubleshoot capture issues
05:45 when Happy Complete Cpmputation
05:55 Notes Turn Green as they c0mplete May take a while
06:12 Double click TEXTURE BUTTON
06:30 Export Model to Edit
06:45 Augmented Reconstruction
08:14 Live Restruction
09:10 All Mesh are Hollow-Plane Cut
09:30 Change Scale
09:43 Sculpting
10:03 Wiki Wiki Time!
step 1 Drag photos into Programme
Step 2 Right Click Structure from motion-Compute
Step 3 Wait until computed
Step 4 When Happy Finish Compution
Step 5 Double Click Texturing Note to View the model
Step 6 Export Model
Step 7 Plane Cut in 3D sotware
Step 8 Change Scale to suit
thanks. Love your comment
thx for the bookmarks realy helpfull
love you
Very helpful thanks
You prolly dont care at all but does anybody know of a trick to get back into an instagram account?
I stupidly forgot the login password. I appreciate any assistance you can give me.
This is unreal technology. Going to be using this for scanning objects for my college game project. Can’t thank you enough for showing this.
This is the most amazing thing I've seen in at least 24 days. Now I don't know if I want to 3d print something or put it in Unity... or whether I want to mod my house into GTA5... or build a rig that automatically rotates the camera around an object.
how can u make mod for gta from 3d file?
@@talharoxen you mean putting the 3d model into GTA? You use the mod tools, I forget which one you use to export to rockstar format but it should be easy to find.
There's 5 or so tools in total that help you mod, it's not as easy as a Bethesda game but it's not that hard thanks to lots of documentation and community support.
@@polydynamix7521 Thanks. I will try to find them. always wanted to make and mod cars in gta
Note: Always take note of what focal length you're shooting at - take not of what lens, focal length, camera type and settings you're using. If your scan comes out wrong it can help you when having to enter metadata manually if a problem occurs. It's especially important to make a note if you're using a smart phone camera because often vital data such as focal length is not recorded in the metadata - which is some cases makes the pictures unusable if you can't figure out the focal length and enter it manually.
I can strongly recommend people using the following trick: Import the photos into a program like Photoshop or GIMP and add a mask around your object removing detail in the background. You can also zoom in if you couldn't get close enough to the object in the first place. Remember to retain all metadata while exporting your photos.
Thank you!
imagine drone preprogrammed photo sequence shots. you can get some nice details of sculptures now.
I did this with a local cathedral (with permission, of course). My results then weren't great however, so I may have to try reprocessing the images with this software.
I'd imagine a cathedral with all it's arches and stuff would also be a very complex model to photogrammetry.
*+Acerbjorn* - Look up "Lofotkatedralen", it's not the gothic type :p
There is a Litchi software for DJI drones that can do that. You just set the waypoints and camera angle and it should work (but be careful if you use that)
@@logitech4873 you either need a better camera or fly closer. If you used Photoscan or even RealityCapture, you will not see improvements in other software.
This is by far the best video about meshroom. There are tips here that Google has a hard time finding. I am impressed.
It has advantages, but also disadvantages. This is why we combine different scanning techniques for perfect results. Great video and great progress in the software!
What other scanning techniques do you normally use?
@@mageenderman Well, I dont work with photogrammetry, but I have read that they also use laser scanning. Its the most precise way to scan a object and the results are outstanding. Here is a video by Konami scanning Schalke 04 stadium and showing previews from the game: ua-cam.com/video/qAkUF4Kbgxk/v-deo.html
I bet you can also take in-game screenshots to effectively "export" asset 3d models from games that dont allow you to access such data.
I tried it and it sucks, game graphics are not that close to reality... :/
@@4WDJUSTY How many shots did you take?
you might be interested in GL Intercept github.com/dtrebilco/glintercept here is one possible application (note this guide is quite outdated): brian-horning.com/google-earth-to-obj-using-glintercept/
I did that with an ingame item from csgo. Worked quite well actually. Used 120 shots or so!
You can also take screen shot of google map (3D buildings) make a model, and sell it for 15 USD (yes some people actually do that)
Going to scan my MK3 and print out a pocket sized version. :P
how did it turn out?
why not just print a new one
He touches on it in the video, but the single most important thing you can do to achieve high-quality results is to use a good camera. Yes, you can get good results with a Gopro, cellphone or point and shoot cameras, but if you compare a model generated from a GoPro to a model from a DSLR/mirrorless camera using a good prime lens, the results from DSLR are absolutely superior.
When comparing models between lower-end cameras and high-end cameras, look at sharp corners/edges, photogrammetric reconstruction with lower-end cameras struggle to produce sharp corners, they come out more round. If possible, use a full-frame camera.
Make sure you move between shots, no standing and pivoting. Take steps in between photos. Take the photos in a systematic way: go low, medium, and high all the way around. For tall structures, you can also scan vertically (take photos as you move up the structure, scoot over a bit then come back down until the whole thing is covered).
You want to disable autofocus as it causes focus changes between photos and will cause pixel blur when the algorithm looks at different photos of the same feature. Set focus to infinity and leave it there.
You want to achieve high overlap between successive pictures, both vertically and horizontally (like 70% to 80% overlap). More overlap means more matched features means a more dense point cloud which means a tighter mesh. If you want to know how good a job the software is doing with your photos, turn the textures off and look at the raw surface. The texture hides lots of artifacts.
Use a fast shutter speed as well to take motion blur out of shots if your shooting by hand and not with a tripod (this is especially important if shooting from a drone).
Use a polarizing filter to help reduce glare if shooting outdoors. Like hard shadows, the software struggles with how to handle it.
Post-process your photos to increase contrast in shadowed areas.
TIP for photogrammetry 3 tutorial - about making whole model of the object - including a bottom (or sole).
If it's a small object you could use armature wire to hold it up so that you could see under it. In any case, try to put the model on a base that is smaller than the model.
Awesome, i know exactly where that first statue is, one of my favourite spots in Prague. I go to the cafe in the National gallery and that restaurant to the right all the time :)
Awesome video. I am loving photogrammetry. My favourite hobby and will definitely give meshroom a try
Great video! Just a small correction: Wikipedia is another website. Many sites have wikis, but Wikipedia is something else. 😉
this looks very promising, can't wait to try it out this weekend!
Did you try it out!! Just curious
Awesome video! Would be nice if you do a live tutorial on 3d reconstruction of more problematic objects, like ones with shiny and flat surfaces without texture. I.e. like reverse engineering old injection molded parts with glossy surface and flat color, or something like this.
Just when my modelling abilities were starting to improve, this comes along and makes modelling obsolete. I've wasted my life. It is rather cool though. Can't wait to try it!
Hey they're not obsolete! For mechanical parts, CAD is still the best option. And for organic shapes, you'll be really happy to know basic modeling when working with photogrammetry meshes, they almost always need a bit of editing/clean up before printing :)
@@Prusa3D It's good to know I'm not worthless just yet lol. It's funny you should mention CAD, as I recently started using CAD to design machines, and have also sucessfully made game assets from CAD models. All of the metal targets in my FPS game are actually real-world products that I produce. The game models are just slightly reworked CAD models.
This is awesome. Planning on getting a prusa very soon.
Thank you for the video. I was new to the topic and tried to read up on it, but your video was just perfect to get into it. Especially the hint that you can make ideal photos on cloudy days has helped - on cloudy days it is not lacking here with me anyway...
Thank you for making this photogrammetry tutorial. Very informative and helped me.
Woooo... I just discovered last month!!! Awesome software!!!
Awesome work. Great tips and I can vouch for many of your tips from my own experience getting Meshroom finally working this week.
Great video!! Thanks for introducing Meshroom to us and a tutorial on how to use it for photogrammetry :) I was going to start doing this soon with Autodesk ReCap, but this might be a good alternative as well.
Oh wow, I've recognized the accent and the statue of Dvorak, then I found the video about your printing farm. Great job you're doing there! I wish I could visit the farm one day and film interview with you!
you're asking basically the daddy of the whole maker scene for an interview :-) he's a legend for sure...
Yes meshroom is great! I tried to use meshroom at work to create a reference model of a complex assembly, but management was too stupid to understand why it was necessary. Let's just say the original documentation was useless for building something with thousands of parts. A 3D model of the entire assembly could've been built with photogrammetry in a couple of days, then distributed to every assembler for reference. NASA did something similar for their space craft where they always built two of everything. Photogrammetry is a much more affordable alternative to this.
Meshroom only works if you have a NVIDIA grafic card.
I you have another one, there is no chance to create a 3 D model.
YA I JUST NOTICE THAT , FUCK
What a time to be alive!
Amazing!! Hello from Brazil!
Hi!
OpenCL support, please.
This calls for an Update.
4:19 - the main thing I learned was that you can graffiti bushes! 😂
Great vid thanks! I'm trying it out now! 😃
Great vid - this gets at what 3D printing is really useful for - graduating beyond Thingiverse downloads.
This was very good explanation of the mesh program I definitely give it a try since it has been updated.
Brilliant! Wow! We’re moving on aren’t we?!
so good for new beginners! thank you!
I hate your teams amazing video editing ability ;) another RAD video guys
Rad
Better than spending £350 per year for autodesk recap pro subscription! Liked and subbed.
This makes me excited to try photogrammetry again.
Me Too... Absolutely!!
This reminds me of the "Building Rome in A Day". Great video! :)
Thanks for explaining this. I have Meshroom already but hadn’t got it working properly yet.
Thanks a lot it's exactly what i was looking for.
This is going to be awesome for my 40k conversion projects
never seen meshroom worked corretly
Awesome, will have to give this a try soon. Great work guys
this is Awesome thank you for taking the time to make this video
great video for more advanced scanning!
Awesome, thank you for your great work!
I really want to do this with an airplane to get a good model for CFD simulations.
I was thinking of using drone for photogrammetry. Years since I did any photogrammetry and it was all manual measurements when I last did it.
We've been doing it since WW2. Probably the most common photogrammetry
@@dudeabides23 - Indeed. I learnt photogrammetry from a guy called Sid Ray, who I suspect had a fair bit to do with military photographic interpretation around that time.
About 11-12 years ago, there was a prototype of this tech. Looks like it reached to it's full potential over the years.
There's still plenty of room for improvement, especially when it comes to speed and ease of use. But yes, it's pretty incredible how fast did the technology improve :)
- Mikolas
For symmetrical objects, you can save time by taking photos of one side and mirroring it in post.
Good tip!
Best photogrammetry tutorial
Glad someone else also loves Dvorak! Got a favorite symphony?
Just as usual, beautiful video, incredibly made
This looks really cool
Very cool. thank you for letting us know about this program. I've not been super happy with photogrammetry since 123d Catch was taken down and this looks like it may get me back in to it again!
Amazing. Thanks for sharing. Just looking for something like this.
6:10 CUDA only >
Same... Killed it for me
Thankfully I have Nvidia
fuck
😢😢😢
Yeah, dont get why developers use something stupid as cuda when there are free open source alternatives that work on any GPU.
Excellent - very much worth a visit.re-visit! Thanks!
Thanks Josef always sharing your knowledge ;D !
Holy cow that is awesome!
great video. i always use photogrammetry but now also use projected light scanning. but the texture maps is the main reason for photogrammetry.
Dannng thoses shoot, so pleasant to watch
Hi Josef, would love to see that 3d sculpting episode! (especially with Blender and 3d printing in mind).
Blender 2.8 all the way!
Cosplayers could be photographed for figures this way. Oooh!
I doubt its possible since the object needs to be perfectly still during photoshooting
Thanks for updating this :-)
Awesome video, very informative
The live reconstruction might be fun if you have some sort of file syncing tool. Take pictures on your phone and have the model already processing by the time you get home.
Thanks for this tutorial. I will give it a try
Have fun! Once you try it, you'll see photogrammetry opportunities everywhere :)
Great video
Thanks! Will try the software again sometime.
Best results are:
1. Record with a videocamera.
2. Use video editor to save video as frames.
3. Use a chroma to minimize false nodes.
What's a chroma?
The traditional way of 3D modeling real-world objects is time consuming, so an automated, easy and affordable way of using photogrammetry that reduces production times where 3D modeled objects are used is a good thing.
Thank you for this tutorial!!!
Love this !
keeping those algorithms happy
How well explained! Subscribed..I would like to see a next video with a drone capture of an urban space.
the WOW at the beginning though
I look forward to prusa’s insight on the new iPhone lidar capability to do this
Thank you very much for your review!
Imagine how far we will have come in 20 years. Conveniently scanning everything within seconds with our "phones", if they are still called like that in the future.
i am amazed
Great Video,Thankyou..
Props to give james maxwell the respect he deserves
Wow, uprime videl sem thumb asi pred 3 tydnama nebo kdy a az ted sem se k tomu zase nahodne dostal a koukam ono to ma pres mega sheldnuti a je to cech :D ... no nic, tohle pokud se mi vysledek podari dostat az do UE4 tak to bude super. Diky moc za info ohledne meshroom!
Hello all, if you are trying to do a really precise close-range 3d scan using photogrammetry, cover gloss, transparent, and homogeneously colored surfaces with texture, such as random lines drawn over a masking tape or a checkered cloth with fine details. Otherwise, the software cannot determine the depth and small details. They do ok but you cannot chase millimeter level detail. Also, when you import .obj files to a "CAD" software, you won't have a texture to help you, just a mesh comes with its faults. If you plan to do some stuff on the mesh, place some marker bumps or lines on the surface to help and guide you. If you know more, you are more than welcome to contribute to this post. Thanks!
Thank you Jo
Finally! Real technology
Some advice on photogrammetry:
1-Use softboxes to illuminate your scene...lights must be flat and casting the least hard shadows as possible.
2- use a solid background..white, black or green..ive gotten good results with blackand very little image retouch.
3- mask or use chroma key on your pictures..this will eliminate all the innecesary information from the image, letting software create tie points only from the points of interest.
4- use a DSLR..really. Most DSLR equipments include important exif data such as focal length, iso, shutter speed and the type of lens used...some of them can even estimate yaw, pitch and roll..which helps most photogrammetry softwares compensate errors in the alignment procedure by using such camera info...effectively tying more points from the pictures overall.
5- use a turntable...keeping your camera fixed in one place and rotating the object of interest can produce amazing results..but due take in account you must forcibly mask your picture...this process can take some time...take in account that there is a point in which taking more pictures will not make a significant difference on the fidelity of your model..you dont need to go up to the fivehundreds or thousands of pictures for something you can easily reconstruct with only 30 to 50 pics. Keep it simple.
6- Postprocess your photos...i recomend using adobe bridge with the camera raw add-on to increase the contrast of multiple pictures simultaneously..this will help software distinguish borders more easily and an overall result in a less noisy point cloud. Just remember to not clip your highlights or crush your shadows too much.
7- Reconstruct in a low or medium quality so you can save time...if you are conviced with the results..simply repeat the process in a higher setting..this will help you save extra time..since medium reconstructions can take only minutes, you can realize faster if the photos you shot were good enough to really reconstruct your model or if they were a complete miss..
8- Play with your settings...overall experiment...photogrammetry depends on multiple factors...photo rig..lighting..the complexity and material of the model..background..image noise..photography postprocessing...keep tweaking until you amount enough experience to realize how to have smoother workflows.
9- You will make mistakes, learn from them...my first reconstructions looked like chronenberg monsters, very little tie point clouds, bad lighting, bad photo quality...you will learn from these mistakes along the way.
Today I can reconstruct models up to a million+ poly count with realistic textures and so can you. The important lesson here is that you learn a long the way and keep trying new things.
Unfortunately, we can't agree with some of those. Especially for Meshroom!
2,3) Do not use background and definitely do not chroma key your pictures, this decreases the quality significantly. We talked about it with the developers: "the results are worse because we will only have points on the object which represent a small sub-part of the image, so the camera parameters are less constrainted which leads to less precise results"
4) EXIF data are included in pictures from phone for almost a decade, we got excellent results just by using smartphone camera
5) No turntable with Meshroom
I love doing photogrammetry. The only thing that I wish was different is that there were more software that runs natively on a Mac. I know Macs are a smaller part of the overall market. Before anyone says "Get a Windows machine!" I'm planning on adding an Nvidia card to my Mac Pro and make my hard drive dual boot to have both macOS and Windows. I just know Mac better than Windows right now.
It was amazing!
For those that dont have Nvidia & don't have to means to get a graphics card. you can use a virtual machine in Microsoft Azure that has it.
it comes about 60cents per hour
I have an idea, with the same setup with the white background, just use a motorized rotating plate in the bottom, and use an automatic shoot every 1 sec
Very cool. Thank you
Great intro
Buen tutorial, didactivo y claro al explicar!
I've tried Meshroom because I wanted to do a tutorial on it but I found it complicated and really had trouble getting good results. These days I use Reality Capture. It's not free but pretty cheap.
Nice. I wish there was a MaxOS version.
I've used PG on old footage from movies or documentaries...
Comparing the models made from old footage with models made of the same scene with newer footage reveals subtle changes in shape.
Besides buildings and construction, it is fun to see a hillside 'heave' during the rainy season and 'shrink' by the end of summer.
Sure, it's only a few feet out of thousands of cubic yards, but is can be measured!
Similar things are seen with Grand Canyon footage; old fly overs and newer ones show lots on interesting 'shifts' in form.
Are volcanologists using this?
Is there any way to use this technology with an older mac? i have a MacBook Pro (13-inch, 2017, Two Thunderbolt 3 ports)
All Photogrammetry programms ive tried wont work with my system.
Very grateful for any tips, or programms that i could use