iModeller 3D Professional Demo
Have been reading about 3d scanning methods recently looking for a lower cost solution then some of the 10k-30k scanner systems and was excited to stumble across the iModeller app. Gave it a try today and was pretty pleased with the results you were able to extract as a first time user. It’s pretty easy to use and has a great price at $800 compared to some of the other solutions. Here’s some screenshots of the test.
You load in a set of images sitting on a flat calibration chart, they recommend using 10-20 shots. I loaded 30 initially, which about a dozen were to low of shots and they were unable to calibrate properly. In the end I ended up using a total of 5 shots for technical reasons I’ll describe below.
You then run the calibration mode. Which will generate reference points for the 3d voxel space based off the distortion of the calibration chart looking threw your specific lens. In my case a 18-200 mm VR2 nikkor strapped onto my d90 with the focal length all the way back at 18mm. The 18mm focal length is introducing a good amount of distortion, which i did adjust for in Lightroom using the distortion profile for the lens. How planar the calibration chart is will affect the accuracy of the model ( read right off the product’s site ) and my chart was showing a bit of waviness, most likely because I printed it last night and the humidity here has been high! You can also manually adjust the reference points which I choose not to along with using a somewhat wavy chart considering the software is in demo mode and I was unable to actually export a model. No need to spend any laborious time making a super accurate model.
After you calibrate your images you run the masking function. The software will attempt to mask the profile of the object in all of your images automatically with a little bit of tweaking needed supposedly. Which is the reason I only used 5 images in the end. For some reason the program would not automatically mask my images, I didn’t investigate why not for very long and just painted a few out manually. The program could use a few more tools for masking in my opinion, like a vector pen tool. But it does have a polygonal line tool which was better then using just a brush. it also took me a bit to figure out how to pan, they don’t offer the standard pan move, instead you have to zoom out then center zoom in. Or paint over the frame edge and drag. In the end, you end up with something like this, which is used for the model point placement generation along with being used for the texture sampling. You can also describe in each image where to be used at, for geo/tex.
After the calibration and the masking you can generate your geometry and the uv mapping. Here you see that the model outputs what you would expect for topology. Not very good geometry for use in 3d animation envioronments as raw data, but pretty great results as geometry for use as a 3d under-drawing which can then be retopologized or constructed in an overlay fasion. The results are pretty amazing in my opinion especially considering it’s constructed for a whole 5 source images. I should also note it appears you can construct larger images then what appear to fit on the calibration chart. If you look on the iModdler product website they have an example of scanning a deer’s head which is on a stand that appears to be up to two feet tall.
Here’s a few screenshots of the actual final model, which should be taken with a grain of salt. I really didn’t spend anytime finessing the setup inside the software as I just wanted to quickly test the software. You can see some nasty texture isssues, which are a result of not enough images used in the calculations along with having some uneven lighting when I captured the source images. I couple probably use some softboxes here!
The UV layout that is outputted is also to be expected, is sort of an automatic mapping that has chunks of samples used from each image all combined into one map. I’m sure with some reading and practice you could get some cleaner maps. But again, the texture output could also just be used as a reference for another means of projection mapping later.
Oh yea and did I mention iModeller also exports right into Cinema 4D?
All in all I was pretty impressed with my quick test results. The models generated may not be the finals, but they would be a very fast way of blocking them out. It’s hard to determine an exact average workflow time just based off this one test but I would assume could have your blocked out model done in an hour. This test took a total of 3 hours, which included setting up my mini-stage. Taking the shots, processing them, fumbling my way through the interface, running through all the procedures, and writing this blog post.
Here’s a link to another modeller out there that use the same Photogrammetry technique. It is a higher cost solution, but also appears to generate a better model along with having more advanced options and mesh/spline output possibilities for cad drawings.