Photogrammetry workflow with Tim Hanson, RealityCapture & CG Society
ฝัง
- เผยแพร่เมื่อ 3 ส.ค. 2024
- Have you missed the CG society webinar? You can now watch the recording online.
Find out:
* why to use photogrammetry software RealityCapture
* 5 tips on capturing the imaginary for photogrammetry
* demonstration of data processing with RealityCapture
Try Demo version of RealityCapture for Free: www.capturingreality.com/Prod... - วิทยาศาสตร์และเทคโนโลยี
![[LIVE] : ONE ลุมพินี 73 วันนี้!! คู่เอก "วรพล vs พันฤทธิ์"](http://i.ytimg.com/vi/rrOljlqse3A/mqdefault.jpg)
Very useful insights. Thanks for the tips !
Excellent summary of your photogrammetry workflow. Your walk-thru is one of the best I’ve heard. A computer-controlled turntable and clear acrylic stand would help a ton when doing repetitive small/ medium object capturing. Iconasys shutter stream 360 software automates the process and makes consistent, accurate image overlap or coverage super easy. Then remove the backgrounds in a batch process for easy export to Reality Capture. The Sketchfab and Epic Games collaboration is wild now too. Keep up the great work Tim 👍🏻
Thank you so much for this video! I've been working with photogrammetry for a while, but this helped me so much to reassess my workflow :)
I guess Im randomly asking but does anyone know a method to get back into an instagram account..?
I was stupid lost the account password. I would appreciate any assistance you can offer me!
@Maxton Trevor Instablaster ;)
@Jad Barrett thanks for your reply. I got to the site thru google and Im trying it out now.
Takes quite some time so I will get back to you later with my results.
@Jad Barrett it did the trick and I actually got access to my account again. I'm so happy!
Thanks so much, you saved my ass !
@Maxton Trevor Happy to help :D
Thanks ! You have inspired me, an interesting issue that did not interest me much before.
this software is a game changer for me.. so incredible! I thought I'd need to buy a 3d scanner for me sculptural projects!
Awsome tutorial! Thank you.
If RC had an in-built retopology function into clean quads which retrieve UVs, would be the perfect software for photogrammetry.
amazing work !
Unity has a delighting tool. It may help with hard light
Thank you for this brief training. however, we would also like to see a thorough, comprehensive training in which the deep, deep features of the program are taught.
Hello Ahmet, thank you for your comment. We are planning a whole new series of tutorials in the future. Stay tuned :)
That's the Korean Friendship Bell in San Pedro. Nice choice of subject! Great video, thank you.
a - How long did Tim take to take the photos on set?
b- how long did Tim take to process the photos and turn them into the final asset?
I wonder for projects like this Lidar would be more feasible?
Amazing!
Hey, great tutorial
I stucked with aligning mesh or orienation. My object need to be aligned with axis, for example i tried to scan a straigt wall or fence, and need it to be straigh aligned to x axis. Of course i could do it later in maya or blender, but need to reduce polycount which is not very convinient in some case. Is that possible to set such alignment or oerintation directly in RealityCapture?
Serious question: Has Depth of Field changed meaning? People talk about 'lots of depth of field' as if it is an artistic effect where the foreground and background are out of focus whereas, historically at least, a large depth of field means the subject is in focus over a large range i.e. exactly the OPPOSITE of how it is being used in this video (and many others).
I noticed this as well. I hope it's not a trend.
"Lot's of depth of field" here means "Lot's of the effects of a small depth of field"
Unfortunate.
It’s something I’ve also noticed. I think the original meaning still stands, but there is a ‘secondary use’ which you note. It’s usually clear what is meant from the context.
Awesome video! One question that I want to be sure on....Are the steps:
1. Import images
2. Align images
3. Use control points with point cloud to align components.
4. Then re-align images again to now merge them together?
5. Then build the model?
Thanks!
The best scan 3d is to NOT use control point. If you have multiple component, the best way is to take photos to complete your scan
Great stuff, thanks for the walkthrough. As someone who has more of a gaming background, could you perhaps use modularity (as in game development) to solve your photogrammetry projects? Instead of taking 4+ photos of each pillar (mo-whatchamacallit), you could do just one and then copy paste it throughout the scene? I know that wouldnt be 1:1 with reality, but say if you wanted to save on production time...? Just asking if that would be technically possible. Cheers!
This particular asset is already fairly modular and would probably only take a day to create from scratch using modelling in-house!
To cut development cost, a low resolution model accompanied with photos could base the above.
This particular example might be better for archeology where subtle variations in the architecture are recorded (I bet the original building isn`t spot on).
He has left out a measuring ruler on it too.
Is there a huge diff using high rez jpgs vs tiff ? Does bit depth make a huge diff? Noticed a comment below about cam type/lenses... Mentioned was "most users use mobile devices" What did Tim use? Canon 1Ds? How does this cam weigh up against an iphone 7plus to a 11pro Is there a way of you illustrating the differences between several devices so it makes it more comforting to decide where one "really" needs to make investments in hardware?
This is gold
F16 on the Fullframe will kill your image to blurry mess (lens resolution limit). I use Sony Alfa R2 and its sucks at F more than 6-8. What kind of DLSR camera you use? I found Nikon do sharp images good for CaptReality.
I can only get one 4K texture, I adjust the settings as he does, follow the steps, unwrap. But the Texturing info still says 1x4092 texture. Help!?
what would be the difference in processing speed and quality in this particular project using 100% quality JPEGs instead of 16bits TIFs?
Hello Pavel, in general, yes JPEGs are smaller and so easier to process. But it also depends on what settings you will use and size and resolution of your files.
This is unbelievable. How did you do it?
So if understood correctly, it's ok to change your distance to the subject if you just keep the same focal length? I mean, going really up close and getting detail in the stone and so on. But would you in this case have to revolve the stone pillars to get the high detail as well?
Hi, there are no restrictions for processing images in RealityCapture, you can take images from different distances, and also use different lenses or cameras. The general rule is to take as many images as you can to increase your project quality.
@@CapturingReality Cheers guys
Awesome. Do you know what the specs of his machine were... wondering why it took 1 day 17 hrs for texturing...
Hello R B, we asked Tim and this is his answer. "Intel dual core 2.6ghz
, 128 gigs of ram, Nvidia Quadro p6000. It mostly took over a day to texture because I was generating over a dozen maps at 16k Rez."
@@CapturingReality Awesome. thanks for the info!
Could I use Matterport scans with this?
superb video,
I want to know how to vectorize the image to make a shape file output
Some CAD software is available which accepts object files.
sooo seeing as you combined the pics with drone footage, when taking pictures of an object (going around the object), should I take some overhead pics as well? should I take some detail pics as well? will it help? I'm just trying to figure out how this works, so I can have a better 'scan'
Yes, you should. You can watch this tutorial for taking images for RC: th-cam.com/video/9e_NLp_FaUk/w-d-xo.html
@@CapturingReality thank you so much!
@@CapturingReality just out of curiosity, how does the software know the position of the camera? Or is it magic?
@@Hejirah It is not a magic, it is geometry, simple speaking. It is something like human eye. You have two of them and therefore you can see the world in 3D.
@Capturing Reality Very interesting video, and guidance through a photgrammetry workflow from start to finish. What camera was used for this project? Any camera suggestions in general ?
Hi, not sure about the used cameras, but in RealityCapture you can use different cameras, it depends also on the quality of exports. Mostly are used different drone cameras and DSLR cameras.
How do you deal with long shutter speed caused by high F stop + overcast(low light) situation? shooting hundreds of photos from different spot with tripod would be just painful. I don't wanna use high ISO either cuz of noise.
Make (or buy) an automated rig.
Hi greetings from Turkey,
I watch from some tutorials that they are suggesting 4-5-8 camera setups that had one cam from eyesgiht othars from groun level and over head and both sides etc. to prevent from parallax. What do you think about that workflow.
Hi Gokhan, what cameras are used there? I think you can also take images that way with one camera, but with this setup it will save your time. If you follow the basic photogrammetry rules, then it should be OK.
Can I put this into Meta quest 3 and walk around it freely ?
Very interesting. I wonder with thos 3d model, can you do measurements? You click one point to another and it will automatically gives you the length width etc? If so my company may be interested to learn more and invest. Please advice. Thank you!!!
Yes, the model can be also measured. There is a tool Define Distance, which can be used also for measurements. But the model must be right scaled to show the correct distance
@Capturing Reality thank you very much. I could not understand your sentence. Could you please rephrase, so I can understand details? Do you sell this software or do you have company? How can I get this software? Thanks!!!
@Capturing Reality I saw you put the link of software for free demo. I wonder photos taken by s20 phone camera will work for measurements with this demo?
@@melihaslan9509 There could be measured distances over the model. You need to place two point there and the distance between these points is computed. But if you want to get the right, real distance, that the model have to be scaled. This means, there are implemented distances or coordinates from real world. You can use the application as PPI (www.capturingreality.com/RealityCapture-PPI) or bought full version (www.capturingreality.com/Products).
@@melihaslan9509 You can create a model with this camera, but as I wrote, the photogrammetry model is basically without scale, so you need to scale it to get the right distances.
Can the cameras be exported like syntheyes?
Would you use a ND filter for mid day drone shooting?
The best conditions are when it is cloudy. But if there is no option for that, that it is possible to use filters.
Is regard3d good for a starter who wants to make games ?
about lens distorsions.. doesnt reality capture do that for you? Ive seen another video mentioning RC knows which lens you use and automatically "fixes" the images during aligning? Im confused
To which part of the video you are referring to?
Yes, RealityCapture computes the distortion parameters during alignment, but you need to choose, which model will be used there.
@@CapturingReality The "adobe lightroom" part at the beginning. I adjust only highlights/shadows/exposure and color temp, I leave the rest to RC. Also: I've seen you are using .tiff files, are the results better than converting RAW files to high quality jpgs? Does it increase processing time?
@@dainjah Can you write a time? I am not able to find it. You can use various formats. I suppose TIFFs should be better than JPG. In that case you can also use the layers workflow - JPGs will be used for alignment and meshing, TIFFs for texturing.
@@CapturingReality 12:40. I just wondered if you used tiffs because they give you denser point cloud (no compression). Anyways, I need to do some testing even with .dng files. I hope my PC can handle it :)
@@dainjah OK, so I suppose you can compute the lens distortion parameters and use those in the alignment process, but it is not recommended to undistord the images before using them in RealityCapture. You can keep that for it.
Regarding the format, it depends on your needs. but those TIFFs could take more time to process. Resulted mesh shouldn't be so different. It also depends on the way how the images were compressed.
Tim, I noticed in a previous tutorial that you are misusing the term "depth of field" and you do so again here. You state you want to eliminate as much depth of field as possible so that the maximum amount of the image will be in focus. In reality, one should have as much depth of field as possible because by definition, depth of field in a photo refers to the distance between the closest and farthest objects that appears acceptably sharp.
can use 2:1 panorama photos?
Panorama images are not ideal for this
@@CapturingReality another question, if I may ask. I do real estate photography. If I take several shots of the floor vertical shooting down from tripod. Can CaptureReality produce a flat image of the floor?
@@360VRStudios Yes, if you will follow rules for capturing images for photogrammetry (th-cam.com/video/9e_NLp_FaUk/w-d-xo.html)
Hey funny question: Do you know your meyers briggs profile? If not can I make a bet that it is INTP?
I don't wanna be a party pooper and i love the software, but there is no fucking way you can call this a production ready model. There are many reasons for people to stick to the 4 week pipeline you've mentioned, at least for now..
Agree. And a talented artist can make such model in 1-2 week alone, not necessary a team. Lots of parts are just repeated. Nowadays a very very complex model can be made in 4 weeks.
@@iani_2020 1-2 weeks? I would finish this in 1 day, 2 days tops. It´s a very easy geometrical model to make
It's not really free dear... need to pay so far.
But the censor sizes are different sizes in camera and drones. The focal length will be different