June 11th, 2020 - Misc

Sidetracking from the current fooding and gardening fun (which I love), a small update (as and insight) on the GemSuite project.

Overtime I showed you guys how a cheap webcam was enough to scan a piece of rough or even an already faceted stone (I went down to 4mm for my first test, with room to spare), use that to generate a 3D model of said stone, and use it to create a tailored cutting design.

One problem that I had left aside when starting that part of the software was parallax error, due to the camera's viewing angle. I've pushed the project so far beyond what it was initially meant to be (a tool for myself), that now it needs to be done with proper precision on every aspect.

I've left that simmering on the backburner for the last locked-down months, with several solutions, from the easiest to the most complicated.

The problem is that, if considering the outline from a single picture, we can't have two perfectly coplanar halves of the outline : there is parallax error.

The easy solution is to take only half a profile at each picture and turn the stone 360 instead of 180. Twice the pictures, twice the work for the user, but way better accuracy of the 3D model.

The harder solution is to keep the current capture mode (180) but use a dedicated calibration dop with a flat rectangle permanently welded to it, representing a perfect viewing plane, and add a pre-calibration algorithm and step to the software to compensate for the parallax error (scanning said dop and having the algorithm compute parallax correction).

The first solution would need very little work on my side to modify the Scanner module, but would lead to doubling the captured images for every stone, hence twice the time for the end user to capture a stone.

The second solution would keep user work almost as it is, only adding a pre-calibration step before scanning a series of stones. On my side, it would need to add another element to the user interface (albeit only in the  calibration  tab), but most importantly to devise and write a parallax compensation algorithm. An added benefit would be that such an algorithm will be needed for the factory calibration of the gem scanners anyway...

I've been playing around in my mind with both solutions for quite some time, while working on the house and preparing my sister's workshop, the garden, the bathroom and sorting out my own workshop (aka huge mess with the bandsaw sawmill still not finished and taking up a decent amount of floor space) at the parents'.

Not yet sure which solution I'd prefer, since the idea is to keep the cost as low as possible to enable the kids that would like to afford GemSuite with very little money (as in $5 webcam, free trial version and that's it), while the dedicated calibration dop seems almost mandatory... I also feel that I have been a part of this community for long enough now to know that the grown-up faceters (>99% of the community) can afford the hardware, and recoup the cost, so I'm kinda leaning towards the second solution rather.

Which of these two solutions seem to be the better, in your own opinion?

Tom, a.k.a. Ludwig Von Sodabowski