08-28-2023, 11:58 PM (This post was last modified: 08-29-2023, 07:21 AM by longvandyke.
Edit Reason: Clarification, truncation, and upload of images
)
Hello,
Why would visuals disappear from an image in Jux upon reloading a parameter file? And is there anything I can do to prevent this or, better yet, restore the original?
I am thinking in particular of the "hypnotic circle" type shapes in some Julia sets. Maybe there is a more appropriate term for them, they're the tunnel-like stacks of interior-colour circles.
I don't understand the math of fractals that well so there are still a number of parameters I fundamentally don't understand, and I don't know of anything in computing that would change if you left the parameters exactly the same. I am using the free version of Jux.
Here's what happened. I used the "classic" formula (z = z squared + c) to zoom in on a promising "minibrot". This was at a zoom level of 12,328 – see Image 1 attached.
I then did a Julia jump to an area where I saw a circle/circles. If I'm not mistaken, this resets the zoom level. I moved around until I found an area where there was a series of hypno circles at different zoom levels. Zoom level was now around 2,995 – see Image 2.
I followed the circles, smaller and smaller, and then zoomed in on an exterior-colour (white) area next to one of them. After zooming enough, inside of that area started appearing a strange cluster of several hypnotic circles. I thought this would be interesting to make some art. I could adjust it to make it look like a character of sorts. At the right position the circles lined up to create a kind of lopsided body and face.
Actually, it went very well. I spent quite some time adjusting lighting and colour. But when I reopened the parameter file, the cluster of hypno circles was gone! This was at a zoom level of 13,402 – see Image 3 (note I had changed the colour scheme by then).
If lost, I'd have to redo the image, but I'd still like to know how to prevent visuals disappearing in the future.
08-29-2023, 07:40 AM (This post was last modified: 08-29-2023, 07:43 AM by Garth Thornton.
Edit Reason: Combining and updating replies
)
In general, parameter files should recreate the image exactly, but there are some exceptions. Those circles are due to the Low Bailout setting. You may need to turn Auto off and adjust it. The problem is that Auto is remembered as a UI setting when you close Jux, not in parameter files. Perhaps it should be in parameter files, but then you would need to remember to turn it on again when needed, so there's no ideal solution. Low bailout adjustment was always intended as a correction rather than a deliberate effect.
I don't know why the exterior is different in the last two images; perhaps that is something you intended.
08-29-2023, 07:59 AM (This post was last modified: 08-29-2023, 08:24 AM by longvandyke.)
(08-29-2023, 07:15 AM)Garth Thornton Wrote: Hi,
you need to upload attachments or insert images or link to a public url, not link to a drive, if my guess is right as to why your images aren't showing.
In general, parameter files should recreate the image exactly, but there are some exceptions. One possibility is that the low bailout needs increasing (turn Auto off) but that usually shows before saving the parameters. The precision and zoom limits for Julia minibrots are lower than for Mandelbrot and could be an issue. I can't tell without seeing the pictures and possibly a parameter file.
Hi Garth – thanks. I uploaded the images at about the same time you responded, I think. Adjusting the low bailout of the file does make circles show again, in the case of decreasing (the file loads at the maximum), but I don't see them returning to the positions in which they had been.
I've directly messaged you with a drive link containing the parameter files for the three images. But I've also added a fourth parameter file, called "Approaching Hypnotic Cluster". Zooming in on its centre shows the cluster of circles I was talking about. (Edit: but Auto low bailout has to be turned on).
And now, I have roughly reproduced the image, and noticed that rendering does not output what is shown in the preview. You can see that in the drive link as well; it is too large to post here.
I did notice that at that point I had zoomed so that the low bailout, set to auto, reached its maximum, and I kept zooming. Could this have had something to do with it?
Edit: Okay, I've read your updated response. The different exterior was a colour change on my part. I think I'm starting to see the behaviour: turning auto off and momentarily adjusting zoom level seems to reveal whether the circles will disappear, and this seems to be connected to having adjusted the zoom level to a point where bailout is no longer visibly increasing automatically. I suppose the innocence toward the purpose of bailout may have contributed to its creative use.
I should probably increase the maximum low bailout. There's no reason you shouldn't use it creatively if it works. Setting it too high for the zoom level can result in noise. Besides getting rid of circles, the usual reason for having to increase it is to eliminate discontinuities in interior coloring (which doesn't always work due to other causes of discontinuities.)
I've played around with the parameter files. Ordinarily you would want to increase max iterations (nice result) but there's no reason to have to. There are some oddities with Julia having very high zooms that I don't understand, such as why adjusting low bailout manually has different results from zooming with auto on, and why the render is different from the preview with some settings. So I have to spend some time on this to see if anything can be done about it, and if not, document it. Thanks for providing the files.
(08-29-2023, 08:50 AM)Garth Thornton Wrote: I should probably increase the maximum low bailout. There's no reason you shouldn't use it creatively if it works. Setting it too high for the zoom level can result in noise. Besides getting rid of circles, the usual reason for having to increase it is to eliminate discontinuities in interior coloring (which doesn't always work due to other causes of discontinuities.)
Thank you, Garth, I'd likely put that to use.
I did notice loading the parameter file, turning off auto low bailout, decreasing it until there's a huge hypno circle, then reactivating auto, had the result of recreating the hypno-character more or less how I remember him. However, the UI then displays the bailout number 25, which is the same as when the project was opened, and the little guy still doesn't render.
I don't know how it works, but maybe the info is useful somehow.
(08-29-2023, 09:37 AM)Garth Thornton Wrote: I've played around with the parameter files. Ordinarily you would want to increase max iterations (nice result) but there's no reason to have to. There are some oddities with Julia having very high zooms that I don't understand, such as why adjusting low bailout manually has different results from zooming with auto on, and why the render is different from the preview with some settings. So I have to spend some time on this to see if anything can be done about it, and if not, document it. Thanks for providing the files.
I will have to increase the limits of Low Bailout, as sometimes it needs to be more than three times the zoom.
The problem with interior coloring is that it needs low bailout to be in an appropriate range for the formula, location and zoom, and there isn't a simple way to automatically determine the range. For Julia sets, critical points are analyzed to determine the convergence characteristics, which are factors in the coloring. The analysis takes max iterations into account (currently using twice max iterations) and generally works well when max iterations is sufficient for most pixels to escape or converge if they are going to. But with max iterations set lower the analysis is not accurate and the relationship between low bailout, max iterations and convergence detection (and the resultant interior coloring) becomes sensitive and erratic, so they don't behave predictably when you adjust them. The reason for the analysis is not just for convergence detection but to allow correct identification of attractor type, period and field angles, to support the various color options.
The problem with the current automatic low bailout setting is that it is too low for Julias of minibrot zooms but too high for deep zooms into simple Julias, while being about right or slightly low for most unzoomed Julias. The problems are most noticeable with interior textures or field lines, which get discontinuous if too low or distorted if too high.