Error submitting photos using 3D objects (beta)

[tonyac55643645]

Get a generic error when trying to submit a photogrammetry job for processing - "An unexpected error has occurred. Please contact technical support via Help > Report a bug..."

In examining the log, the error is a File_ReadError: "File read error. Please check network, storage and access rights. ("C:/MyPath/011627.png").

I've tried several things to address this:

• I've confirmed the files are located on my system drive (same drive as the program).
• I've moved them up in the folder hierarchy thinking perhaps a long filepath was the issue.
• I've renamed them to only include numbers thinking that special characters in the filenames were the issue.
• I've tried using fewer images (I took a total of 240 images - 48 angles from 5 different elevations).
• I've tried running Sampler with admin priviledges.
• I've tried importing masks as well as allowing Sampler to mask the images itself.
• I've tried resaving the files with a different color space because the log had referenced an issue with the sRGB IEC61966-2.1 color space.

Despite all this, the error persists.

Any help would be appreciated. I have the files available, but they would far exceed the attachment limits. I'm happy to provide them on request.

[Cyril Dellenbach]

Hi @tonyac55643645,

Thanks for the details.

Having a lot of pictures is a great thing. Substance Sampler needs a decent amount of images to work properly, and the more you have, the better. You just have to make sure the dataset doesn't exceed 6G pixels (6 000 000 000 pixels) in total, which is the software limit. It represents 500 photos of 12M pixels.

Could you share a bunch of photos from the dataset (10 would do the job), so I can check there's nothing obvious preventing the reconstruction? If there's not, I'll probably ask you for the full dataset and your log, so we can make tests on our side.

Regards,

[tonyac55643645]

Thanks for getting back to me.  Does the pixel limit include masks or not?  I'm under the pixel limit with just the photos (about 4.3G pixels), but the masks would obviously double that.

I have tried allowing Sampler to generate the masks itself and the error persists, so I suppose it's not the masks.

Here's a link to a Google Drive folder with 10 photos and associated masks:

https://drive.google.com/drive/folders/1yRMg_7jpJjCGihzeditokSOkakErMCKi 

[tonyac55643645]

I masked the photos myself because Sampler and Photoshop were both having trouble with the decals on the softball.

[Cyril Dellenbach]

I think the limit doesn't include masks, but I'm not 100% sure to be honest. I'll check with the dev. In the meantime, it's worth trying without masks.

Considering the shared pictures and the mentioned error in the log, I'm confident the problem doesn't come from the dataset. Could you send your log at cdellenbach@adobe.com ? We'll take a closer look at this issue.

Regards,

[Cyril Dellenbach]

I just realized you also sent a bug report, so I already have access to your log. No need to send another one.

[tonyac55643645]

I had tried letting Sampler generate masks before, but I gave it another shot just now.  Same end result with same error.  Now the log includes a list of mask generation errors for all the photos like this:

[WARNING][Debug: ThreeDCapture.cpp:361 Alg::ThreeDCapture::ThreeDCapture::generateMasks::<lambda_b7e730d730987a6d26f2ffff2cd8d326>::operator ()()] Error while creating mask for photo :file:///C:/MyPath/005-621.png: no_subject_found

Considering the subject is pretty dominate in all of these photos, I'm guessing the issue is related to the file access error.  Photoshop had no issue identifying the subject, although the selection wasn't perfect.

[Cyril Dellenbach]

I advise you to even try without masks. That being said, would it be possible to have the full dataset to investigate on our side? You can send it at cdellenbach@adobe.com 

Let me know.

[tonyac55643645]

Just emailed you a link to the entire set.  It may go to your junk folder.  Please let me know if you don't receive it.

I tried Sampler just now with no masks, but same result.

[Cyril Dellenbach]

I was able to try the dataset on my end. You have a clear data set, consistent focal, sharp images and no reflects. However, I wasn't able to do the 3D capture either. In my opinion this is one of these situations justifying that the feature is still in Beta.

Maybe that the background color could be a bit more similar to the turntable (Sampler is easily triggered by a static background with a rotating object), but that shouldn't prevent the 3D capture.

If that's okay for you, we'll keep the dataset for further tests.

Let me know if you rather not.

[tonyac55643645]

Please do.  Thanks for looking into it.