Superb video, thank you; I’ve been trying to get some 3d models of parts from some model kits that I need to replicate, but my results have been just awful! I’ll give some of your tips a go, but I’ve only got a D3200 with the standard lens, not a macro lens (my Galaxy S9 produces better pictures at such extreme close up!) and focus and depth of field are the biggest problems. I’m not a gifted photographer either, so my knowledge of shooting in anything other than Auto is limited. I had read that I should use Aperture Priority mode with a low ISO so that tallies with what you say. Anyway, very informative and helpful, thank you very much
@@3Dforensics Even the most difficult object has a way. It might take 17 tries. But everything is possible. Remote control or code is required for macro scans absolutely.
don't forget to turn off autofocus on your video camera setup. ;) Something is throwing it off, causing it to pulse. That's hard to do if you're solo of course, but you have a wide DOF anyway, so even if it's not Quite in focus no one will complain. I've enjoyed your tips. I'm only using a sony a6000 with meshroom, because I'm just doing it for fun and that costs nothing, but always have troubles getting things to work well. Hopefully with all these tips I can get better results, which will hopefully speed things up and give me better results. So far, for most things I get better results with just my iphone 12 LiDAR, so I obviously have some learning to do. Questions for speed v quality like should I keep the images raw? shoot wide open, or keep it as iso100, use a tripod and use a longer exposure. Longer lens gives you more detail because you get more resolution per shot, and less distortion, but that's Way more shots needed, so where is the balance for speed v quality? These experiments take time. Thanks for answering some of those through your tips.
Thank you Andy, much appreciate the comments. Experimentation is where you get to make a lot of mistakes with the reward of learning and that is what I do a lot of...failing and learning. Just today I did an experiment with a 360 camera that was a complete disaster!
Agisoft way of masking seems better with it´s "from Background" & "from model" method. Is 3D planing something similar ? It usually takes me a very long time to create masks in Zephyr. The masking is the only reason why I prefer Agisoft. Anyway, could you create a video where you show how to improve the overall quality by using filters such as as "X84" & "reprojection error" in Zephyr. I really enjoy your content, keep up the good work!
There is an automated masking method in 3DF Zephyr that can be quite handy but you should have really good control of the background, otherwise, you’ll need to do a bit of touching up but I have had some success with it. I will have a look at your suggestions for future videos. Thank you!
@@3Dforensics Thank you for the quick reply. I have tried the masking feature in Zephyr and while it does work well in certain situations it´s not quet precise. One more thought regarding future videos. Maybe you can show Zephyrs "photoconsisrency" function ? This function seems to be powerful, but I don´t really get how it works and what it really does. (Does it really reproduce more details acc. to the pictures taken ? Or is it producing " false details" where it reads certain color information as deepend or raised from a surface ? Like a tattoo could be read as 3d information by mistake by the software) Please don´t pay attention to my grammar. I am from Germany and english is my 3rd language.
Well, I think it depends on what you want. If it’s a fully comprehensive photogrammetry package, then PhotoModeler is excellent. If you are mostly interested in 3D reconstructions, then the top 3 are definitely MetaShape, 3DF Zephyr and Reality Capture. Any of those will give you good results but I have become very proficient in 3DF Zephyr and I absolutely love their masking tools!
@@scottkarlins I saw Context Capture presented at a conference when it was rather new. It looked like it had some good results, but at the time, the cost was a bit of an issue for a lot of users. That may have changed nowadays but nonetheless, Context Capture may be a good solution for some users.
So all this complexity in the video to say that you took 10 pictures of a thing with a macro lens and were able to create a relatively decent 3d model of it?
We did a project where we were reconstructing ridge details on the finger and when you get to very small sizes, it is quite challenging. Try it. There are issues that become apparent pretty quick.
Yours are by far the most thorough and informative photogrammetry videos on youtube. Thanks for putting these out, they’re REALLY helpful.
Thank you, good job. 👍
Superb video, thank you; I’ve been trying to get some 3d models of parts from some model kits that I need to replicate, but my results have been just awful! I’ll give some of your tips a go, but I’ve only got a D3200 with the standard lens, not a macro lens (my Galaxy S9 produces better pictures at such extreme close up!) and focus and depth of field are the biggest problems. I’m not a gifted photographer either, so my knowledge of shooting in anything other than Auto is limited. I had read that I should use Aperture Priority mode with a low ISO so that tallies with what you say. Anyway, very informative and helpful, thank you very much
The object you are trying to reconstruct matters too! Keep at it and thank you!
@@3Dforensics Even the most difficult object has a way. It might take 17 tries. But everything is possible.
Remote control or code is required for macro scans absolutely.
Been using notes for years because of the s-pen capture feature. Now on the s22 ultra and it's amazing for phone field capture
don't forget to turn off autofocus on your video camera setup. ;) Something is throwing it off, causing it to pulse. That's hard to do if you're solo of course, but you have a wide DOF anyway, so even if it's not Quite in focus no one will complain. I've enjoyed your tips. I'm only using a sony a6000 with meshroom, because I'm just doing it for fun and that costs nothing, but always have troubles getting things to work well. Hopefully with all these tips I can get better results, which will hopefully speed things up and give me better results. So far, for most things I get better results with just my iphone 12 LiDAR, so I obviously have some learning to do. Questions for speed v quality like should I keep the images raw? shoot wide open, or keep it as iso100, use a tripod and use a longer exposure. Longer lens gives you more detail because you get more resolution per shot, and less distortion, but that's Way more shots needed, so where is the balance for speed v quality? These experiments take time. Thanks for answering some of those through your tips.
Thank you Andy, much appreciate the comments. Experimentation is where you get to make a lot of mistakes with the reward of learning and that is what I do a lot of...failing and learning. Just today I did an experiment with a 360 camera that was a complete disaster!
Agisoft way of masking seems better with it´s "from Background" & "from model" method. Is 3D planing something similar ? It usually takes me a very long time to create masks in Zephyr. The masking is the only reason why I prefer Agisoft. Anyway, could you create a video where you show how to improve the overall quality by using filters such as as "X84" & "reprojection error" in Zephyr. I really enjoy your content, keep up the good work!
There is an automated masking method in 3DF Zephyr that can be quite handy but you should have really good control of the background, otherwise, you’ll need to do a bit of touching up but I have had some success with it. I will have a look at your suggestions for future videos. Thank you!
@@3Dforensics Thank you for the quick reply. I have tried the masking feature in Zephyr and while it does work well in certain situations it´s not quet precise. One more thought regarding future videos. Maybe you can show Zephyrs "photoconsisrency" function ? This function seems to be powerful, but I don´t really get how it works and what it really does. (Does it really reproduce more details acc. to the pictures taken ? Or is it producing " false details" where it reads certain color information as deepend or raised from a surface ? Like a tattoo could be read as 3d information by mistake by the software) Please don´t pay attention to my grammar. I am from Germany and english is my 3rd language.
thank you but this model is to easy you should try with jewelry.. that's real challenge..
Definitely more challenging!
Do you have any videos or opinions on different reconstruction softwares? I tried a bunch and found Agisoft's solution to be the best, is that true?
Well, I think it depends on what you want. If it’s a fully comprehensive photogrammetry package, then PhotoModeler is excellent. If you are mostly interested in 3D reconstructions, then the top 3 are definitely MetaShape, 3DF Zephyr and Reality Capture. Any of those will give you good results but I have become very proficient in 3DF Zephyr and I absolutely love their masking tools!
Don’t forget Bentley ContextCapture!
@@scottkarlins I saw Context Capture presented at a conference when it was rather new. It looked like it had some good results, but at the time, the cost was a bit of an issue for a lot of users. That may have changed nowadays but nonetheless, Context Capture may be a good solution for some users.
So all this complexity in the video to say that you took 10 pictures of a thing with a macro lens and were able to create a relatively decent 3d model of it?
We did a project where we were reconstructing ridge details on the finger and when you get to very small sizes, it is quite challenging. Try it. There are issues that become apparent pretty quick.