Agisoft Metashape

Agisoft Metashape => General => Topic started by: johnnokomis on December 07, 2022, 08:24:18 AM

Title: New Import Depth Image feature
Post by: johnnokomis on December 07, 2022, 08:24:18 AM
Being able to import photos with a depth map image in the meta data is what I've been asking for since I first started using Metashape. It works and it works very well, however, it has a few things I'd like to bring to the devs attention hopefully.
I'm using an LG V60 with the ToF sensor. The only way I know how to capture the depth data needed is by using the portrait mode on either the stock camera app or a GCam app that's been ported over from the Pixel phones. This mode does some heavy post processing on the main photo captured. My V60 has a top notch manual mode with great results but no way to use it and capture depth data with it. This isn't Agisoft's fault by any means but something I had to mention.
What could be done differently is during the import process though. I've found that if an object is too close to the ToF sensor at the time the image is taken, the depth data isn't saved but the normal photo is. You have no way of knowing which photos in your dataset have a depth map along with it or not. So when importing a large group of photos and a few mixed in didn't capture the depth data... Instead of just not importing those few images it refuses to import any of them. Good luck sifting through the 100's of photos to find the few rotten apples.

I love the new feature and I'm not complaining a bit. I'm just wanting to help make it reach it's full potential.

Thank you!
Title: Re: New Import Depth Image feature
Post by: Kiesel on December 07, 2022, 10:56:22 AM
Hello Johnnokomis,


Yes an improved, error less prone, import of photos with depth maps would be nice.

I have found this for extracting depth maps with Exiftools in 'Jay's Technical Talk':

Quote
exiftool original_image.jpg -trailer -b > depth.jpg

So it should also be possible to sort photos without depth maps out before import in Metashape . Btw. I have found that a too fast photo shooting can lead to missing depth maps as well.

There is also a web page, which can show and extract all layers of a Portrait.JPG https://www.photopea.com/ (https://www.photopea.com/).

In my own tests with my Google Pixel 4a 5G, which doesn't have a ToF-sensor - only a dual pixel sensor design, I have found that the depth maps of it are unfortunately that bad for low textured surfaces that this added no advantage over normal processing. I think this is where a ToF-sensor makes a difference. Would be interesting to see some results from your LG V60.

Best regards,

Kiesel