Hi there,
i am just validating your softwarea and i can’t get a functional alignment of my images. Of about 50 images (360°) of a simple boxy furniture it only can match 5 or 8 images. The Software only recognizes similar images from the same view. i.e i get a component for the front view and another component for the side view in a complete different orientation.
But i dont get a decent 3d result.
Is there a a software setting i am missing, because the images are mostly fine. Other softwares were able to produce a result.
Hi sbe
Can you upload the data for inspection ?
Hi Wishgranter,
thank you for taking time looking into the data.
We took a new set of images and it gives us a better result, though still not a complete 360* alignment.
here a quick image
and the data we used: https://we.tl/ty4mhNEden
A good number of the images are totally out of focus. That is bad for photogrammetry and you should try to avoid it. Maybe use a tripod?
Another tip:
Put some different colored post-it notes on the floor around the object to help with the alignment process.
thanks for your reply.
We tested it right now with markers and with approx 160 images. The results were pretty decent. On a second test with the same markers and only 51 images, the software wasn’t able to compose the components. Is there a rule of thumb to how many images are needed for the software to find enough match points?
It depends on the complexity of the object, how many details there are, shiny/reflective surfaces = bad, unicolor surfaces = bad, repetitive surface patterns are also bad…
But it also depends on your camera.
lowest ISO possible to avoid noise, Semi-high F-stop for good depth of field, more mpixel = fewer photos needed, better detail, …
Here is an example of an object that gives you full alignment and a complete 3d model (top, bottom) in just 36 photos:
https://dl.dropboxusercontent.com/u/25588164/Chest.zip
It works because there is a lot of features for RC to latch on to and align the cameras.
I found out that putting newspapers and magazines all around the objects is good too.
Anything that add some texturing and details to the floor will definitely help the alignment process.
thanks everybody for those suggestions.
we received better results for the same chair. now we have troubles with a table and we are not quite sure why we do not get a clear aligned component. We compressed the images for uploading a bit.
Can you have a look at the images we used for the table. Would help us a lot!
https://wetransfer.com/downloads/975034 … 341/8b1167
ShadowTail wrote:
It depends on the complexity of the object, how many details there are, shiny/reflective surfaces = bad, unicolor surfaces = bad, repetitive surface patterns are also bad…
But it also depends on your camera.
lowest ISO possible to avoid noise, Semi-high F-stop for good depth of field, more mpixel = fewer photos needed, better detail, …
Here is an example of an object that gives you full alignment and a complete 3d model (top, bottom) in just 36 photos:
https://dl.dropboxusercontent.com/u/25588164/Chest.zip
It works because there is a lot of features for RC to latch on to and align the cameras.
Hey ShadowTail,
nice whitebox! 
I just tried it out and it is simply fantastic!
Top and bottom snap together as if it floated in space.
I guess the whitebox helps enormously…
Wish I could put my buildings in one!
Did you shoot @ 12 mpx?
Raw or jpg? (my old question)
The model looks already quite good in Normal, only a bit orange skin and some artefacts, but I guess in High it would be perfect!
Thanks for sharing!
The camera I use can’t do more than 12 mpx.
It is, after all, just an entry level canon eos 1100d (Which I am planning to replace somewhat soonish).
I do jpg most of the time. Its more than good enough for what I need.
Nothing wrong with 12mpx! Me too. 
Just another good example for good results with reasonable equipement!
