I thought I have a 3D the printer that is capable of printing multiple colour. If I could use it to make a Calibration target it will be interesting.
After importing the original PDF from PhotonVision to Inkscape I was able to convert the pattern to SVG, then using SVG 2 STL I was able to make a STL for the pattern. After importing it to the slicer and assemble it with a square base plate and some colouring, I got my sliced file.
The print quality come out pretty good. When the printed part remain on the print plate, it remain flat (which is good for calibration), and the scale seems ok.
The only way to ensure the flatness is to have it remain on the print bad and use the print bed as a carrier. Print bed tends to lose a adhesion overtime and not everyone had a spare print bed. I want to try putting some double-sided tape and fix it to another flat plate oo see how well it hold.
just in case if multicolour printing is impossible, I want to try use a sharpie to colour the whole thing black and see how well it work.
It’ll use more filament, but why not print it on top of a 8mm 15% gyroid infill rectangle? (Or whatever settings will give you enough strength with your filament, slicer, and printer…)
A manual filament swap should also work fine for this. I know PrusaSlicer makes it pretty easy to add a pause between layers to switch out filament for multicolor print.
I’ll add to the list as I don’t understand how this can work with the commonly available calibration calculations. With the slight height of the squares, only straight-on are the dimensions correct and actually everything will be slightly off center and will see the side of the height as a change in width. I’ll be interested to know how well these calibrate compared to a flatter piece of paper. In this age of fussing about 1/2 a pixel I’m guessing these won’t work right.
There are (expensive) 3-D calibration targets and I’m sure they would require different calculations.
It can be made on a 2d printer but the goal of this is for it to be very flat and most 2d printers can only take in a flexible paper like material that does not come out flat. You could tape that paper to a flat board like I have in the past but this is to try to make a better solution that does not suffer from the paper stretching and not becoming flat on the board anymore.
The current cal recommendations talk about making sure the target is very flat. We owe a docs update to make this more specific, but the best practices we’ve found so far (in order from best to worse) are:
A matte laptop screen or modern computer monitor (the manufacturing process causes a pretty solid guarantee of flatness)
A purpose-made calibration target with a nice stiff backing
Printed target, measured and double-checked for accuracy with a calipers, and spray-adhesived onto a backing that is known to be flat.
[throws on mentor hat]
This is because the underlying calibration logic, currently, is hard-coded to assume all black/white quad corners lie in a 2d plane. Making them not in a plane (especially a consistently warped plane) has the effect of throwing off the lens distortion model coefficients, with (anecdotally) very unpredictable results.
Sometimes it’s obvious, sometimes it might just be inducing a random few-inches error in your pose estimate. We’ve yet to figure out how to predict goodness here.
I’d have concerns with a 3d print over warping (esp. for a full-bed size print), and the fact (at least my) PEI textured bed is visibly less flat than any of the above recommended solutions.
[throws on PhotonVision Dev Hat]
A 3d printed target wouldn’t show up in the recommended list until it could be shown same-as-or-better-than one of the options on the list… and I’m not entirely sure how to prove that for all team’s 3d printers. Or even for 1 team’s 3d printer.
Due to how glass is made, it tends to be quite flat while being relatively cheap. Importantly, it doesn’t like to bend. At all.
Metal and plastic don’t really do the same. (acrylic is brittle but bends a lot more than glass does)
I’m not sure about whiteboards which would be my other go-to idea.
Print it out on a standard printer, cover the back with glue, and use a paint roller to get it on there good.
Alternatively, you could straight up etch it in the glass if you have a decent laser cutter.
