Difficulty scanning in miniatures from web sources

[Eternal3lade]

So I'm having a bit of trouble when it comes to scanning in miniature figures using 36 photos (12 photos at three different angles). In this case the source images are from the below site. (They have a 360 spinning photo section) Basically nothing aligns and the sparse cloud is all over the place.

https://www.games-workshop.com/en-WW/Sicarians-2017

My typical workflow is to extract the images from the virtual turn table, upscale using topaz gigapixel or imbatch (x4), import into zephyr, mask the background using the color picker on white and process them from there. I've scanned hundreds of these and maybe have a 50% success rate from this specific method and photo configuration.

Is there anything I can do to improve my success rate? I typical use max settings in the project wizard, incremental engine, no bonding box, and approximate grid. Maybe there are some advanced settings I need to be looking at?

If it helps, I can attach my scaled images for folks to take a look at.

[cam3d]

Hi Eternal3lade - Welcome to the forum!

As 3DF Zephyr works using feature detection on a pixel level it's not recommended to upscale images before input (even if in some cases it works, it's generally not beneficial) - Your source imagery is fairly weak (low number of images, low resolution, third party source) so there is always going to be a likelihood of failure. Some subjects will have more features for Zephyr to work with, and therefore increase the likelihood of success despite the non-ideal datasets.

A setting you could try adjusting would be using 'unordered' rather than 'approximate grid', but ultimately you need to take your own images to boost your chances of successful camera orientation and surface reconstruction.

[Eternal3lade]

Thanks cam3d ! That's about what I thought, I just wanted to verify my thoughts. When I scale the images up using Topaz Gigapixel I think I'm actually creating more features that Zephyr can utilize as illustrated by the below "scaled" image comparison. I suppose that's why I get much better results after scaling, the software is actually doing a bit more than that. If I don't go through this process and only use the base low resolution images I think my success rate is pretty much 0%. At any rate I'll give unordered a try with various settings and see if I can't make some magical success.