3 ms·
Looks very interesting. What would you say are the main technical challenges in the way of your development? By the way, the roadmap link on the Github repo pa
by MITSardine 2y ago
Looks very interesting. What would you say are the main technical challenges in the way of your development?
By the way, the roadmap link on the Github repo page is broken (though I did find it in the project's website).
I saw you're interested in handling Bezier splines. This might be premature generalization, but would you want to support whole BREPs? Is that something that is used in applications? I imagine lenses are generally simple geometries, so perhaps not.
- fouronnes3 2y agoThanks, I've fixed the links. Honestly, the main challenge at this point feels like finding the time / resources to just work on everything I've got on the roadmap. I think I've got enough ideas for at least 1y of full time work, perhaps more. Bezier splines are interesting. I actually had a working implementation a few weeks ago, so I know it will work. It's just that I refactored the internals of how surfaces work and didn't port the old bezier spline code yet, but it shoudln't be too difficult. I haven't given too much thought about volume representation so far. Currently it's just a list of unconnected surfaces which seems sufficient for now.
- MITSardine 2y agoOh, BREPs don't necessarily mean closed surfaces. They're just trimmed rational Bézier splines and other primitives stitched together with some topology. I'm mainly wondering if you might not find the surface-related functions you need in a CAD kernel like OpenCascade, which might support the surfaces you're interested in out of the box.
- fouronnes3 2y agoI see. Yeah I could look at that but it needs to be re-implemented in pytorch to provide the backwards pass anyway. But the implementation might be a useful resource.