Question about masking in Photoshop

[KenmoJr]

Using the FREE version of 3DF Zephyr I tried scanning in this 1/24 plastic scale body of a 1932 Ford coupe.

I am using a turntable and a black back drop, a Nikon D7000 DSLR and Nikon 50mm lens. Shooting in manual mode.

Using Photoshop I created the mask using the "document name" +_masked.png.

Should I be masking out the far side of the body which appears through the open wheel and window areas? Or leave them unmasked as in my example. I've marked the areas with a red arrow to clarify.

Cheers & many thanks

Ken

[cam3d]

KenmoJr - This subject will be very hard to capture with photogrammetry as the surface is primarily homogeneous and feature poor - Rather than masking out the entire surroundings, I'd recommend using a feature rich base (like fine newsprint or similar) which stays fixed in position relative to the subject.

The feature rich base acts as an anchor to help guide the camera positions - you can then mask out the surroundings which move relative to the subject + base so as to keep the input data consistent and avoid disparate features pulling your camera positions off course. I hope this makes sense, let me know if you need clarification.

[KenmoJr]

Thanks for your reply. I fully understand what you are suggesting. I will try it.

However I was going to flip the car body onto it's roof so I could capture the underside to show it is hollow. The Body only has a left & right side and a top, the front and back have openings. There is no floor as it is a separate object.

How would you suggest I then photograph the bottom to inform 3DF Zephyr the underside is hollow?

Cheers & many thanks

[cam3d]

KenmoJr There aren't too many options to capture a subject like this. The best results will come from dusting (which I know is something you would rather avoid) and as a last resort you could combine two scan projects together however this is only an option in the full version of 3DF Zephyr and regardless, will likely provide subpar results on account of the surface type.

If you got inventive with how you mount the subject on your turntable you might be able to capture enough of it to get most of the model constructed, but be careful to ensure it doesn't drift in relation to the feature rich base and pull itself apart digitally with inconsistent reference features across the photoset.

Alternatively you could construct the exterior of the model, then use this as a template to extrude thickness in a third party program, however this will likely require manual intervention and is not something I'm particular proficient in so am unable to advise further.

[KenmoJr]

Well, I'm starting to think photogrammetry may not be for me. Too many variable and considerations. Perhaps a hobbyist type of 3D laser scanner would be a better solution. I've watch Jay Leno's Garage on youtube and in a few episodes he has used his industrial strength 3D scanner to help him fabricate an antique car part that is now longer available or being manufactured. I recall he used the technology to manufacture a replacement timing cover and also an oil pan for an antique Duesenberg and Jay never had to put any powder or dust on his Duesenberg parts.

And it appears I am very much alone in my desire to 3D capture model and real life cars as there appears to be no one else here has any desire to do so.

And I still uncertain if the areas with the red arrows should be left as is or masked out.

I have way too many questions and very few answers.

Cheers & take care...

[cam3d]

KenmoJr - Photogrammetry has inherent limitations which mean it's not suitable for every type of material. The laser scanning systems Jay Leno has showcased over the years such as the NextEngine or Faro Arm are much better suited to the kind subject matter you are looking to capture. This is especially true as you do not want to prep the surface with features, which is a requirement for successful SFM and MVS processing of otherwise homogeneous feature poor surfaces. Different horses, different courses!

Regarding the areas masked out or not, it depends on what you want to reconstruct. If it's masked, it will be ignored, if it's not masked, it will be taken into consideration - It's personal preference when it comes to which parts of the subject you want to dismiss.

[KenmoJr]

cam3d - Thank you very much for your assistance and responses. I do appreciate them. I've decided to give RealityCapture a try and despite it's horrid interface I'm getting much better results from it using the very same datasets. It appears to be more "friendly" with shiny metal surfaces and does not reject as many photos as Zephyr. And has "control points" to help align rejected photos. However I really much prefer the workflow of Zephyr.

However RealityCapture seems to have a much more active forum and bigger user base so it is much easier to find solutions and advice for problems and issues. And the free version of RealityCapture does not have the 50 photos limit.I wish there was a photogrammetry app that combined the best features of RealityCapture and Zephyr.

I've totally given up on Zephyr for now. Perhaps I will return to it. I will check on future releases of Zephyr to see if they've caught up with RealityCapture (photo number limit and dealing with shiny metal).

Warmest regards from Nova Scotia, Canada.

[KenmoJr]

I've been using the demo version of Metashape 1.8 for 28 days. I am getting much better results from it then Reality Capture and Zephyr. And I do not get near as many photos rejected during allignment as I do with Zephyr or Reality using the same photos.

[cam3d]

KenmoJr As a general rule, if your photos are being rejected in any software, it's important to understand why that is the case so as to improve your data acquisition approach, however if you have found a solution which works for you and your images, that's great! Happy to help if you have any more 3DF Zephyr related questions.

[KenmoJr]

Yes, I do have a question. Why does Metashape accept photos that Zephyr rejects? If there was a problem with my photos then Metascape should reject them as well?

[Andrea Alessi]

Not really, while the principle of photogrammetry software are the broadly same, they are different software made with different technologies.

You will always find some dataset or acquisition techniques that will work better with one package and other datasets/techniques that will work better with another.

Try changing your settings to match the needs of the photo that you take. For example, and this may sound counterintuitive, try lowering your SfM settings, as if the photos are badly shot or if the subject is not very textured, it-s better to have less, stronger keypoints, rather than more, weaker keypoints.

[KenmoJr]

SfM settings? I do not know what you mean.


Here is a good example below of a small snowman object I shot with my Nikon D7000 and 35mm lens using manual settings that were at the same settings through out the dataset.
I shot 25 photos. Metashape alligned all 25, Zephyr only 13 of 25.

It has been my experience that Metashape always has a higher acceptance then Zephyr. So I do not think it is my photos.

I like Zephyr very much but I find it is a little too finicky and the support for it is very lacking.

And I am very concerned that in order to get a reply one has to ask the question many times, which can be very frustrating.

Thanks kindly

Kenmo

[KenmoJr]

I downloaded Zephyr 6.5 and gave it a go with this dataset. Zephyr 6.5 only accepted 11 of 25 photos.

[Andrea Alessi]

You may need to change your settings depending on your photos. It is highly recommended you read trhough the documentation and tutorials as stated before, the SfM settings may need adjustments for badly shot photos or difficult subjects. There is no one-fits-all setting. Other package may be more set to accept bad photos and perform worse in other scenarios. With Zephyr you have the full control over the pipeline but indeed, it requires you to change the settings. If you are looking for a one click solution then we are probably not the solution for you, as Zephyr is a highly configurable software that can be adapted to all scenarios, but requires knowledge in limit
cases.

SfM means structure from motion, and is the very first step in the pipeline. You can find tutorials here https://www.3dflow.net/3df-zephyr-pa...-tuning-guide/

here are some more in depth explanation https://www.3dflow.net/3df-zephyr-pa...-tuning-guide/

If you need premium support, consider purchasing the full version of 3DF Zephyr, which grants you faster response time over our official support system. The forum is not the official support vector, even though we reply very well within 24h most of the times, or you can reach out to one of our official trainers https://www.3dflow.net/certified-trainers/