Requesting assistance figuring out photomatching

count23

So i've done a fair few models, usually with blueprints that are semi reliable, like for cars and such. but i've never really built a model from scratch using just perspective photos alone and i'm having some troubles figuring out the right pipeline. I've tried using perspective photos that are as close to orthographic angles as possible and they've turned up bust, i've tried 3/4s views and can't seem to line anything up.

I wanted to expand the skillset to practice modelling without using blueprints while trying to maintain a good level of accuracy with the subject model andi'm stumbling out of the gate on it.

I know most photos have barrel distortion and will never 100% match, and the reference material i have isn't exactly something that can be turntabled (it's a scifi space ship miniature, photos came from various sources so a mix of focal lengths and such too).

I'm hoping someone might offer me tips on how i can get started on this, or a pipeline on how i'd pull off modelling in this scenario.

I was trying to model this miniature, assuming the sphere structure was a proper rounded sphere and not any other deformed state as the closest primitive i could identify as a standard size. So i've been dropping a circle as a start and trying to level the plane with the medium axis of the sphere before working my way outwards, but i've not really been able to make any decent progress: https://www.thetrekcollective.com/2015/06/behind-scenes-photos-of-uss-pasteur.html

So any tips or guidance would be appreciated. If i'm on the right track and it's purely keep going until i get 80%ish accuracy on all photos and just live with it, so be it, but i feel like i'm going about this all wrong.

EDIT: I shoudl add i've tried fspy and perspective plotter but they seem to be intended for archvis more than these kinds of objects, so the results have been iffy at best and useless at worst.

okidoki

Since there are some blueprints.. i would use this as main reference and then use the others "to look at them very closely".. every tutorial using a front side and top view never tells you that the experienced modeller are used to "see in 3D" even if looking at a 2D image. So they "just can do this".. in fact even a experienced modellr looks at dozen of detail photos (or at least should..  ).
Something like fspy woudl give you a proper camera postion to use this to compare your model looked through this camera with any photo. But this "hardware" model is not shown in any environment you can use as a guide. The last one is the most  promising but with the board in the background does not give enough info to extract a proper perspective.. i guess.

count23

most of hte blueprints i've found have been far too inaccurate, i tried to create off them wiht the assumption that again the sphere was a perfect sphere so i could get some sort of basic angle matching for the photos but they just seem so far off it felt almost like a waste of effort even blocking off the blueprints, but if that's still the better way to tackle it...