I am pretty sure that you're aware of this but you forgot to mention at the time of filming this video, that using or experimenting with a polarizing filter to get rid of first light bounces to the sensor from shiny objects such as glass or metal otherwise it was a great overview. Thanks. :)
Thanks for the explanation. I'm pretty familar with 3D applications and have been playing around with photoscanning for the past couple years, but still haven't quite gotten the hang of using my camera, so occasionally I get a set of completely bunked up series of images.
Well, I figure I don’t waste your time with asking to subscribe, donations, sponsor videos or any breaks so the at the very least a nice intro can’t hurt.
Thanks for this. I wish I'd seen it sooner. Some of the stuff I had discovered for myself and can confirm. But there's other stuff like changing the ISO setting I've never played with. (That's about to change). A camera ap' for my phone; I hadn't thought about that before. There have been so many times when my phone was the only camera to hand. I've now seen two of your videos and think they're rather good, (hence the subscription!) Cheers.
Dear Teacher Eugene, I am very grateful for the lesson learned in this video. Everything became very clear to me as a beginner in Photogrammetry for clarity and didactics. I confess that I made 4 out of 5 mistakes, less about shiny or transparent objects. However, there is one detail that I'm still not sure about in the camera's settings: the focus type. I have a Nikon D5100 and would ask you, please, for a suggestion so I can move forward. Before start to learn Photogrammetry, I have used my cam in auto mode and exercised with A priority only. I'm working with a twelve-inch statue, with some details. With the camera's Manual mode, I set AF-S as Focus Mode; Single-point as AF-area mode; Center-weighted metering as Metering. I am sorry if its makes no sense. Please your feedback will be invaluable.
Hi Dimitre. I shoot almost entirely in manual mode when doing a small object where I can control the lighting and focus. This way, your shots are more consistent in exposure in case your object varies in texture. Choosing the high f-stop gives you the greatest depth of field. Are you still not able to get your object in focus?
Sir, Thank you so much for your time and great mistakes that we all might make. I have been beating my head against the wall in trying to scan a 1/16th scale car body that is about 12" long. I have made many mistakes starting from the body being white on a white table with a white background. I have tried my cell phone and my DSLR @ f/11 and 1/250. The car is now flat grey on a manual turntable with a random pattern on it. Still not getting a point cloud, getting really poor texture models. I have tried walk around and with tripod. Now i'm going to try and mist some black paint on it to break it up more. I have also played with lighting. Seems i get good shots late in the day under my patio but still bad results. Never would of thought i had done 3 of your 5 mistakes with number 5 being the biggest one. One question i have, after i do three circles of the part- low, middle and high should i take the 3rd shots of it that you mentioned with the shoe example? I have used up to 60 pictures but still find huge gaps in the point cloud or the texture is very broken. I'm using meshroom and open the .obj in blender. Thanks for all your help.
Hi Bob, if you are using a turntable, then you will need to mask out the background or you will have all kinds of issues. Do me a favour and feel free to contact me offline through my contact form on my www.ai2-3d.com website and I would be more than happy to look at the images and see what we can do.
@@3Dforensics If I may, I have a similar project with similar failures. I have a very detailed character study of my favorite model from school shot on a green screen turning her, rather than shooting around her. Mid chest up, bust composition. No matter how much time I spend in Corel masking and removing the green screen, meshroom places the camera points that it does use, from the actual side I’ve shot her from pointing towards the green screen position. Then, the lions share I’ve shot from the side and back of her head are dismissed by the program. The limited rendering it does produce is useless as is. Are these shots going to end up being unusable?
It depends on the situation. I imagine you’re in a very small space that you want to capture and in that case, a wide angle lens would be much more beneficial. On the other hand, drones don’t need a wide field of view since they can just fly higher.
Hi and thanks for the video. What software do you use to process the photos into a mesh? I have been trying to use Meshroom with mixed results. I am using a turntable with a vertical arm no the side to get a full range of photos but annoyingly some photos are not accepted, no reason given. Any insights you have would be much appreciated.
@@3Dforensics I have just started playing with Zephyr today after watching another of your videos. These software packages are a bit out of a hobbyist price range (double the price in Australia) but I know now that I have to improve my photography. Once again thanks for sharing your knowledge and I'm going to watch more of your videos. Regards.
Hi There, thanks for such a clarifying video. I've got a question, I have a DSLR Nikon camera, and iPad pro M1. I want to create some digital models of some LEGO pieces. I want to export this to a CAD software later on, so I need to use IGES, STL, SHAPR, DWG or DXF files. Any recomendation what software could support this? Thank you so much!
Creating CAD models isn’t my strength as I would normally just model this directly in something like Rhino or other software. However, I believe that 3DF Zephyr has some “detection” for planes and such like that but to be honest, you will likely find that when producing a photogrammetry model of flat and shiny pieces like lego, the surfaces don’t reconstruct all that well unless they have some texture. You could try to coat the pieces a little bit with AESUB or dry shampoo. This should help significantly. Also, I would think that lego pieces could just as easily be modelled in Rhino with a set of calibres and taking some measurements. Hopefully I understand what you are doing properly.
Hi man, may I ask wish software do you use for photogrametry? and if it works to take close pictures of the object where not the entire object appear in the photo, I mean if you are making a car you take a lot of 360 of the car but I don't know if you have to go closer and take only parts of the car like bumpers and tires
Wouldn't getting close to the object introduce distortions? Like if you take a photo of someone's face and get too close, their nose will look stretched out because of how light projects onto the camera sensor. I heard people say that you should get further away, but then zoom in, so you get a flatter projection of the objects and therefore less distorted image. I'm a complete noob in photography, so I would be super interested in the real answer!
Most photogrammetry programs calculate the camera and lens characteristics so it is always best to have a fixed focal length and the more organized the photos, the better. The point is to fill the view and get close when you can. If you have to be further back, that is ok but be consistent and keep the focal length(Zoom) constant.
Is it possible to turn the object over to get all sides when scanning or is it better to balance it on a rod or something like that? I know with a laser scanner you can pause the scan, turn the object over and continue scanning. Great video!
I have a small business that does photogrammetry and other 3d things. I typically use a studio with a turntable and tripod. Problem I'm having with a few objects is getting all the underside angles because the turntable is in the way.
You can turn the object over and get all sides. Please have a look at some of my videos on masking with 3DF Zephyr. Masking is a great tool when turning an object over on a turntable.
So something like Zephyr doesn’t require the whole object to be in every shot? I thought it did.
3 ปีที่แล้ว +4
they don't... sometime that's not even physically possible, when shooting stuff like architecture interiors, also, often you want to go closer in for more detail, even with small items
It really depends on your system and to be clear, what software are you using? Many software packages will use your graphics card to accelerate the reconstruction and if you don’t have the right graphics card, you will have to wait. The number of photos that you use will have a big impact on how long the reconstruction takes as well.
In your first point, about getting in close, you say it's about the number of pixels your subject takes up; but then you say using a telephoto lens doesn't help. This seems contradictory. I suspect it's something to do with parallax, but I was hoping you could explain further.
I often know what I want to say but the words come out all wrong. What I mean to say was that getting in close is not the same as using your zoom lens. You don’t really want to have all kinds of different focal lengths mixed up in a project. If you were to only use a telephoto lens and not use anything else, then that would be perfectly fine. However, there are some subtle differences about a wide angle lens close up to a subject vs a telephoto lens far away from an object. This might be a good episode for Click 3D. Thanks for the following up and hopefully it wasn’t too confusing.
@@3Dforensics Cool. It makes sense that having different foal lengths makes things harder for calculating. Is that mitigated by the camera recording the focal length at all? Instinctively I want to take a few overview shots, and lots of detail shots, but what makes sense to me doesn't necessarily make sense to the algorithms. I was involved in this field a quarter century ago, and have just jumped back in on a wave of nostalgia. So I'm currently trying to teach myself about modern photography ("make sure you've updated the firmware on your lens" -- WTF???) ... whilst learning to operate the camera remotely ... whilst learning to operating a quadcopter ... whilst standing in a muddy field surrounded by sheep. I've had a few false starts -- who knew a 40/150 zoom lens would be too heavy for the quadcopter gimbal? Or that auto camera setting would over-expose all the detail of my subject, and the "repetative structure filtering in meshroom would strip the rest - giving me a lovely model of the landscape around the subject, with a big hole in the middle.... But I'm having fun...
@@jocramkrispy305 having fun is what makes it all worthwhile! With respect to the focal length, the camera does record the approximate setting but it is rarely accurate for photogrammetry. The EXIF data is often used for the initial setup but after that the algorithm takes over to find a more accurate solution.
@@3Dforensics my students would love to know more about your "journey" into a unique career! We do a touch of photometry and 3d printing...thanks again!
Hey there, I'm a fan of these videos but you are very... very verbose. That's ok, but it can be hard to disseminate when you're bringing up a new point from something you're repeating. My suggestion is: consider each of these videos as two parts. Your 'TL:DR' should come first. The FIRST part of the video, sum up everything within a single minute or two. Part 2: get into more detail. Many people will be able to take what they need from PT1 and move on. Those that are more new to the subject will stay for the duration. Yesterday I was with a photographer who I showed this video to and he was literally banging on the 'skip 5 sec' button over and over, like he was playing pinball, trying to land on the ISO/Fstop/ShutterSpeed 'formula' which was buried in the explanation as to what each of those things were (something every seasoned photographer knows). Any good tutorial shows you what you're going to be attempting to achieve before you get started and I recommend you do the same. This dialog: "The five mistakes are 1. Not being close enough to the subject, 2. Not getting enough pictures, including overlap, to fully cover your subject 3. having blurred images, 4. not following the camera settings formula which is 'Low ISO, High Fstop, fast shutter speed' and 5. simply choosing bad subject matter. Now, I'll cover each of these in more detail... let's get into it." and then chapter things off for ease of jumping. Remember that everyone can re-watch areas they want to better comprehend so leave the repeating to the reviewing, not in your script.
Hey man, I’m still new to this TH-cam thing even though I’ve done a lot of videos, it’s not my main gig. Appreciate the comments and will see how things evolve here.
Fantastic video, thanks dude 👍
I am pretty sure that you're aware of this but you forgot to mention at the time of filming this video, that using or experimenting with a polarizing filter to get rid of first light bounces to the sensor from shiny objects such as glass or metal otherwise it was a great overview. Thanks. :)
I would say that isn’t necessarily a mistake, but a good tip!
The gold standard best camera to photogrammetry tutorial I’ve ever sat through. Thank you very much for presenting this.
Thanks for the explanation. I'm pretty familar with 3D applications and have been playing around with photoscanning for the past couple years, but still haven't quite gotten the hang of using my camera, so occasionally I get a set of completely bunked up series of images.
Keep at it and practice with other cameras or even your mobile phone camera. The more you do, the better you will get!
@@3Dforensics ironically I sometimes get better results with my pixel phone then I do with my wife's Sony a6400
@@SleepyMatt-zzz Likewise. Depending on the situation and object, it can be easier with a mobile phone.
This vid help me A LOT! Thanks!
37 seconds for the intro ? Nice video - I was just starting preparing my students to use photogrammetry.
Well, I figure I don’t waste your time with asking to subscribe, donations, sponsor videos or any breaks so the at the very least a nice intro can’t hurt.
@@3Dforensics quite interesting to see how far 3D photogrammetry got
Now it is build into MacOS too.
Thank you so much for your tips. Everything on point. Subscribed!
Thanks for this. I wish I'd seen it sooner. Some of the stuff I had discovered for myself and can confirm. But there's other stuff like changing the ISO setting I've never played with. (That's about to change). A camera ap' for my phone; I hadn't thought about that before. There have been so many times when my phone was the only camera to hand.
I've now seen two of your videos and think they're rather good, (hence the subscription!)
Cheers.
Dear Teacher Eugene, I am very grateful for the lesson learned in this video. Everything became very clear to me as a beginner in Photogrammetry for clarity and didactics. I confess that I made 4 out of 5 mistakes, less about shiny or transparent objects. However, there is one detail that I'm still not sure about in the camera's settings: the focus type. I have a Nikon D5100 and would ask you, please, for a suggestion so I can move forward. Before start to learn Photogrammetry, I have used my cam in auto mode and exercised with A priority only. I'm working with a twelve-inch statue, with some details. With the camera's Manual mode, I set AF-S as Focus Mode; Single-point as AF-area mode; Center-weighted metering as Metering. I am sorry if its makes no sense.
Please your feedback will be invaluable.
Hi Dimitre. I shoot almost entirely in manual mode when doing a small object where I can control the lighting and focus. This way, your shots are more consistent in exposure in case your object varies in texture. Choosing the high f-stop gives you the greatest depth of field. Are you still not able to get your object in focus?
Not yet. I take the photos with the camera in vertical position in a tripod.
I can not see the whole statue in focus. Just the center or the top.
Sir, Thank you so much for your time and great mistakes that we all might make. I have been beating my head against the wall in trying to scan a 1/16th scale car body that is about 12" long. I have made many mistakes starting from the body being white on a white table with a white background. I have tried my cell phone and my DSLR @ f/11 and 1/250. The car is now flat grey on a manual turntable with a random pattern on it. Still not getting a point cloud, getting really poor texture models. I have tried walk around and with tripod. Now i'm going to try and mist some black paint on it to break it up more. I have also played with lighting. Seems i get good shots late in the day under my patio but still bad results. Never would of thought i had done 3 of your 5 mistakes with number 5 being the biggest one.
One question i have, after i do three circles of the part- low, middle and high should i take the 3rd shots of it that you mentioned with the shoe example? I have used up to 60 pictures but still find huge gaps in the point cloud or the texture is very broken. I'm using meshroom and open the .obj in blender.
Thanks for all your help.
Hi Bob, if you are using a turntable, then you will need to mask out the background or you will have all kinds of issues. Do me a favour and feel free to contact me offline through my contact form on my www.ai2-3d.com website and I would be more than happy to look at the images and see what we can do.
@@3Dforensics If I may, I have a similar project with similar failures. I have a very detailed character study of my favorite model from school shot on a green screen turning her, rather than shooting around her. Mid chest up, bust composition. No matter how much time I spend in Corel masking and removing the green screen, meshroom places the camera points that it does use, from the actual side I’ve shot her from pointing towards the green screen position. Then, the lions share I’ve shot from the side and back of her head are dismissed by the program. The limited rendering it does produce is useless as is.
Are these shots going to end up being unusable?
@@jamesmyers2087 drop me a link of the photos through my website and I'll have a look and let you know directly if there is any recourse.
The 1/3 shots are not required if you have plenty of coverage.
Hello Bob,RealityCapture Meshroom Metashape which is better?(I use RTX 3060 12G)
Great information, thank you very much….
Hello,RealityCapture Meshroom Metashape which is better?(I use RTX 3060 12G)
Thanks for those tips. Excellent tutorial.
How about the lens angle? Which is more efficient, wide angle vs. tele objective?
It depends on the situation. I imagine you’re in a very small space that you want to capture and in that case, a wide angle lens would be much more beneficial. On the other hand, drones don’t need a wide field of view since they can just fly higher.
Hi and thanks for the video. What software do you use to process the photos into a mesh? I have been trying to use Meshroom with mixed results. I am using a turntable with a vertical arm no the side to get a full range of photos but annoyingly some photos are not accepted, no reason given. Any insights you have would be much appreciated.
I typically use 3DF Zephyr or MetaShape.
@@3Dforensics I have just started playing with Zephyr today after watching another of your videos. These software packages are a bit out of a hobbyist price range (double the price in Australia) but I know now that I have to improve my photography. Once again thanks for sharing your knowledge and I'm going to watch more of your videos. Regards.
@@garrygiddings7455 There is a version of 3DF Zephyr that is only $200 and they also have a free version that you can use up to 50 photos.
@@3Dforensics Hi Eugene, I'm still at a loss why my photos are being rejected, could I possibly send you some files to look at?
Great video man seriously
Hi There, thanks for such a clarifying video. I've got a question, I have a DSLR Nikon camera, and iPad pro M1. I want to create some digital models of some LEGO pieces. I want to export this to a CAD software later on, so I need to use IGES, STL, SHAPR, DWG or DXF files. Any recomendation what software could support this? Thank you so much!
Creating CAD models isn’t my strength as I would normally just model this directly in something like Rhino or other software. However, I believe that 3DF Zephyr has some “detection” for planes and such like that but to be honest, you will likely find that when producing a photogrammetry model of flat and shiny pieces like lego, the surfaces don’t reconstruct all that well unless they have some texture. You could try to coat the pieces a little bit with AESUB or dry shampoo. This should help significantly. Also, I would think that lego pieces could just as easily be modelled in Rhino with a set of calibres and taking some measurements. Hopefully I understand what you are doing properly.
Hi man, may I ask wish software do you use for photogrametry? and if it works to take close pictures of the object where not the entire object appear in the photo, I mean if you are making a car you take a lot of 360 of the car but I don't know if you have to go closer and take only parts of the car like bumpers and tires
Wouldn't getting close to the object introduce distortions? Like if you take a photo of someone's face and get too close, their nose will look stretched out because of how light projects onto the camera sensor. I heard people say that you should get further away, but then zoom in, so you get a flatter projection of the objects and therefore less distorted image. I'm a complete noob in photography, so I would be super interested in the real answer!
Most photogrammetry programs calculate the camera and lens characteristics so it is always best to have a fixed focal length and the more organized the photos, the better. The point is to fill the view and get close when you can. If you have to be further back, that is ok but be consistent and keep the focal length(Zoom) constant.
Nice vid! A Good Variable ND filter helps cut reflections and glare… :).
Is it possible to turn the object over to get all sides when scanning or is it better to balance it on a rod or something like that? I know with a laser scanner you can pause the scan, turn the object over and continue scanning. Great video!
I have a small business that does photogrammetry and other 3d things. I typically use a studio with a turntable and tripod. Problem I'm having with a few objects is getting all the underside angles because the turntable is in the way.
You can turn the object over and get all sides. Please have a look at some of my videos on masking with 3DF Zephyr. Masking is a great tool when turning an object over on a turntable.
So something like Zephyr doesn’t require the whole object to be in every shot? I thought it did.
they don't... sometime that's not even physically possible, when shooting stuff like architecture interiors, also, often you want to go closer in for more detail, even with small items
@ Gotcha. Thanks.
@Amari Lorenzo You two bots again...
Correct!
Hi,I don't know why the software very very slowly to create the model , please have you idea to solve that problem
It really depends on your system and to be clear, what software are you using? Many software packages will use your graphics card to accelerate the reconstruction and if you don’t have the right graphics card, you will have to wait. The number of photos that you use will have a big impact on how long the reconstruction takes as well.
In your first point, about getting in close, you say it's about the number of pixels your subject takes up; but then you say using a telephoto lens doesn't help. This seems contradictory. I suspect it's something to do with parallax, but I was hoping you could explain further.
I often know what I want to say but the words come out all wrong. What I mean to say was that getting in close is not the same as using your zoom lens. You don’t really want to have all kinds of different focal lengths mixed up in a project. If you were to only use a telephoto lens and not use anything else, then that would be perfectly fine. However, there are some subtle differences about a wide angle lens close up to a subject vs a telephoto lens far away from an object. This might be a good episode for Click 3D. Thanks for the following up and hopefully it wasn’t too confusing.
@@3Dforensics Cool. It makes sense that having different foal lengths makes things harder for calculating.
Is that mitigated by the camera recording the focal length at all?
Instinctively I want to take a few overview shots, and lots of detail shots, but what makes sense to me doesn't necessarily make sense to the algorithms.
I was involved in this field a quarter century ago, and have just jumped back in on a wave of nostalgia. So I'm currently trying to teach myself about modern photography ("make sure you've updated the firmware on your lens" -- WTF???)
... whilst learning to operate the camera remotely
... whilst learning to operating a quadcopter
... whilst standing in a muddy field surrounded by sheep.
I've had a few false starts -- who knew a 40/150 zoom lens would be too heavy for the quadcopter gimbal? Or that auto camera setting would over-expose all the detail of my subject, and the "repetative structure filtering in meshroom would strip the rest - giving me a lovely model of the landscape around the subject, with a big hole in the middle....
But I'm having fun...
@@jocramkrispy305 having fun is what makes it all worthwhile! With respect to the focal length, the camera does record the approximate setting but it is rarely accurate for photogrammetry. The EXIF data is often used for the initial setup but after that the algorithm takes over to find a more accurate solution.
What did you use for the transition effects? haha do you use OBS?
In this video I believe it was StreamYard
If i may ask, has anybody experimented with polarising filters while doing fotogrametry on shiny surface? Does it help?
Yes, it helps in many cases. If I can get some time in the near future, I will work up a video!
Sorry if i missed this in one of your videos.....where are you located...i see a PEng of Ontario Cert. :)
Just north of Toronto in Vaughan
@@3Dforensics my students would love to know more about your "journey" into a unique career! We do a touch of photometry and 3d printing...thanks again!
@@RobotoForgoto contact me through my website. ai2-3d.com
@@3Dforensics cheers!
2:03
Hey there, I'm a fan of these videos but you are very... very verbose. That's ok, but it can be hard to disseminate when you're bringing up a new point from something you're repeating. My suggestion is: consider each of these videos as two parts. Your 'TL:DR' should come first. The FIRST part of the video, sum up everything within a single minute or two. Part 2: get into more detail. Many people will be able to take what they need from PT1 and move on. Those that are more new to the subject will stay for the duration. Yesterday I was with a photographer who I showed this video to and he was literally banging on the 'skip 5 sec' button over and over, like he was playing pinball, trying to land on the ISO/Fstop/ShutterSpeed 'formula' which was buried in the explanation as to what each of those things were (something every seasoned photographer knows). Any good tutorial shows you what you're going to be attempting to achieve before you get started and I recommend you do the same. This dialog: "The five mistakes are 1. Not being close enough to the subject, 2. Not getting enough pictures, including overlap, to fully cover your subject 3. having blurred images, 4. not following the camera settings formula which is 'Low ISO, High Fstop, fast shutter speed' and 5. simply choosing bad subject matter. Now, I'll cover each of these in more detail... let's get into it." and then chapter things off for ease of jumping. Remember that everyone can re-watch areas they want to better comprehend so leave the repeating to the reviewing, not in your script.
Hey man, I’m still new to this TH-cam thing even though I’ve done a lot of videos, it’s not my main gig. Appreciate the comments and will see how things evolve here.