Leaderboard

Here is the result from an improved version of my algorithm, producing the same 750 circles with 250k attempt cap. It now does more accurate occupied tracking, by only marking grid cells as occupied when they are actually occupied (as determined by at least one of the four corners of a grid cell being within or on the circumference of a circle). Unfortunately, it still uses a bounding box template for a circle to be placed to determine whether it can be placed. This is due to the fact that the number of placement attempts, in total, is quite high and the determination of the actual cells that would be occupied quite expensive, relatively speaking. Therefore, it is only done when a circle is placed rather than during attempts of "can it be placed here." Now at least there is more space available to place it('s bounding box). First the new times of five runs: 750 circles have been generated in 2.97 seconds 750 circles have been generated in 6.31 seconds 750 circles have been generated in 4.33 seconds 750 circles have been generated in 3.03 seconds 750 circles have been generated in 2.68 seconds Below is one of the images produced showing less wasted space, although one has to bear in mind that as each of the circles got placed, its bounding box was used to test for overlap with areas already occupied by (slightly exaggerated approximations of the locations of) existing circles. This resulted in some placement attempts being deemed invalid because the corners of the bounding box of a candidate circle overlapped existing geometry, but if the algorithm was smarter it would see that the circle itself did not. I will think about this more and see if I can come up with a way to improve on this and possibly get rid of the bounding box testing altogether, without greatly sacrificing the run time per test.

I wish you could toggle all the modes with W then. Something in Preferences to turn it on so you get all the options, or it restricts it to World/Local (default on). But you know what I just realized and tried - I right-clicked on "orientation" and chose "Add to HUD." And it adds a little menu button in the middle of the screen and you can click it and change directly through that. And it stays when you change components/shading views/etc. I'd love a hotkey but I can't complain anymore now. Definitely helpful while modeling. Cheers!

All 750, for sure, are generated. That is the actual real count. If there is a failure to generate (i.e., max attempts exceeded at some point), I would get an error message indicating how many actually got generated. That is how I determined that 750 was pretty much the limit, at least with my prior implementation. This one might be able to edge closer to 1,000 circles.

Wow ! Hard to figure out how you did that. Looks like a sky animation used in the reflection channel of a torus. But the symmetry of the stars breaks after a while from clouds as if the were mapped differently from the rest of the sky. Well done.

There is a hotkey for changing between World and Local space for transform tools (W), but not, AFAIK to switch the temp modelling axis between its various alignment options. I would suggest the tool presets system here but it is no quicker than just manually changing that one drop down from Normal to Axis or whatever. CBR

Its VRAM with Error Correction Code (ECC). Practically this prevents errors while loading and handling big datasets. Example: when rendering animations for Liebherr with 48 Mio polygons per scene with Redshift an RTX 3060 TI/ 8 GB VRAM failed to load the scene while an RTX A4000H/ 8GB ECC VRAM (mobile) flawlessly succeeded with that. Of course that can also be a question of the driver. But in general ECC VRAM is more reliable in critical, highly complex scenarios. The Liebherr animations were created on an RTX A6000 (48GB ECC VRAM).

That's an extremely misleading statement. It is trained using copyrighted material, illegally , without which it couldn't amount to very much, or at least would take much longer to amount to something. It is only able to do this because image art is not very well protected compared to musical art, which is never used to train AI because it is so much better protected.

A new one - twist deformer prototype! This can be expanded into included one and even improved. In Viewport OM Strength control i nAM Graph is simple 71_Twist_Deformer.c4d