OK, this is getting ridiculous

[GRSnyder]

This is a bug report / gripe...

This document has only seven components (if you count the top level), and to a first approximation, Fusion 360 has colored them all the same color! Well, OK, three colors: creamsicle, lime green, and (one solitary) faded puce.

For the love of all that is holy and useful, could Fusion please not reuse colors until the palette is completely exhausted? And could the palette colors not be so similar to one another that it's impossible to distinguish them unless they are directly apposed in the UI? The autocoloring feature is wonderful, but these minor nits completely destroy its value from time to time.

There's quite a bit on the web in regard to palettes of distinguishable colors and algorithms for generating such palettes. The current palette generates nice displays, but I'd be happy to sacrifice some of the color harmony for an increased range of lightness and saturation. Readability uber alles.

[jodom4 · Accepted solution]

Hey GRSnyder,

That's a good suggestion, I'll mention it to the UI guys. If the colors are too similar, or are repeating too frequently, you can right-click on a component in the browser and select "Cycle Component Color".

[jeff_strater]

I need to look into this,  As far as I remember, there are 10 fairly unique colors, that are used in order.  So, for 7 components, you should not see any duplications.  Can you share the model here?  I'd like to take a look at it to see if there is something weird going on.

[GRSnyder]

---

@jeff_strater wrote: As far as I remember, there are 10 fairly unique colors, that are used in order.  So, for 7 components, you should not see any duplications.  Can you share the model here? 

---

Sure, it's https://a360.co/2rjHkCD. There are a few derived components included, but I marked them public as well.

However, I'm not sure you really need to poke around - I think I can guess at what's going on given your description. This is a model with a fairly complex history (currently at v117). It's a dual fan mount with cooling ducts, and it's been through a lot of major revisions based on CFD studies and real-world experience. There used to be a bunch of components used only for constructing compound joints. Several components have been pushed out to files of their own and replaced with derives. So it's entirely possible that this is just the luck of the draw: the components that happen to remain just happened to draw similar colors based on their position in the history.

I think the main issue is that Fusion 360 never seems to reassign colors on its own. But I'm not sure that's a bad decision - it allows this kind of problem but has a low surprise factor. I hadn't been aware of the Cycle Component Color command (thanks for mentioning that, @jodom4). If I had known about that, I would have just fixed the colors and not come crybabying over here.

That said, I do think the palette could be improved and enlarged. There are several pairs that aren't readily distinguishable.

[jeff_strater · Accepted solution]

@GRSnyder - you're more or less correct in your understanding here, especially in the "that this is just the luck of the draw".  How this works today is:  when this is switched on, Fusion just goes through the list of components and assigns colors from a fixed list of 10 colors.  That's the total algorithm.  At least as far as I can tell, in your design, this does not repeat any colors.  There is no attempt to sort the colors by geometric or browser proximity to "similar" colors.  

I can see that there are some concerns with the color palette.  For instance, in your example, there are two "greenish" components that can be hard to distinguish from each other:

Ideally, I agree - Fusion should provide a richer experience in this feature - including a larger palette of colors, ability to edit or extend this palette, ability to change the assignment of colors, etc.  To be honest, this was never intended to be a complex capability - it was just a quick way to be able to show where the component boundaries lie.

[jeff_strater]

I stand corrected:  There are now, actually, 36 different colors for component color cycling, not 10.  I'd forgotten that we increased that number.  Of course, the more colors you have, the more likely that you will have colors that are pretty close to each other, so it's definitely a tradeoff here.  There just are not that many colors that are easily distinguishable to the human eye...

[GRSnyder]

@jeff_strater wrote: At least as far as I can tell, in your design, this does not repeat any colors.

---

I'm not 100% certain exactly how this works, but since I posted a live link, I think you may be seeing the current version of that file. This version reflects the fact that I went back after learning about Cycle Component Color and manually fixed the colors to make them more distinct. At the time of the original post, the colors were extensively duplicated as shown in the OP.

Just out of curiosity, I went through and compiled the actual palette that Fusion 360 seems to be using on my Mac. (I seem to be missing a few, but whatever...) Of course, color is protean and exact values can shift when captured by screenshot and then edited into a palette and converted to sRGB. Nevertheless, this palette pretty much matches my subjective impression of "Fusion 360's favorite colors":

I think this pretty much speaks for itself: At least six different indistinguishable versions of muted orange. Four indistinguishable versions of icy blue. Four indistinguishable versions of dark orange. I'll be generous and claim only four indistinguishable rosy pinks, but the last color in the rosy pink aisle is probably not distinguishable from the muted oranges. Someone really likes orange. 🙂

After looking into this, I think you (@jeff.strater) are right that there are only so many colors you can include in a palette while maintaining distinguishability, especially if the colors are represented by tiny slivers displayed at distances from each other. And I don't mean that in a vague "surely there must be a limit" sense - the number seems to be about 15.

So, I suspect it was probably a mistake to go from 10 colors to 36. 36 likely can't be done. But 16 is probably doable. And yes, better that colors be obviously and clearly duplicated than that you are left with color assignments that might or might not be the same.

Here's an alternative palette from https://medialab.github.io/iwanthue that only took a couple of minutes to generate:

I'm not proposing this as the palette; it's not as harmonious as the current one. But at least the colors are both distinct and neutral enough to use in the context of Fusion 360. It's just proof of concept that a graphic designer could likely come up with something much better than the current palette.

[jeff_strater · Accepted solution]

This is the source of the dilemma for us:  "So, I suspect it was probably a mistake to go from 10 colors to 36.  36 likely can't be done".  The problem is:  At a low number of colors, in a complex assembly, the chances that you get repeats of components which are physically close to each other goes up.  That was the complaint that induced us to bump up the number - it reduces the chance of adjacent components having exactly the same color.  But, the tradeoff, then, is that a larger number of colors increase the odds of the "almost the same color" problem.  So, clearly we can't ever win...  We are bound to make some portion of users outraged.

Regarding the colors chosen, agreed that there are several colors here which are pretty close.  Here is the actual definitions of the colors used:

<ColorCycleTable>

    <ColorCycle name="Light Purple" RGB="0.89 0.57 0.90" />

    <ColorCycle name="Light Pink" RGB="1.00 0.69 0.70" />

    <ColorCycle name="Dark Yellow" RGB="1.00 0.85 0.48" />

    <ColorCycle name="Turquoise Blue" RGB="0.39 0.87 0.89" />

    <ColorCycle name="Green" RGB="0.77 0.96 0.44" />

    <ColorCycle name="Orange" RGB="1.00 0.62 0.40" />

    <ColorCycle name="Blue" RGB="0.36 0.55 0.93" />

    <ColorCycle name="Light Red" RGB="1.00 0.39 0.45" />

    <ColorCycle name="Purple" RGB="0.75 0.55 0.95" />

    <ColorCycle name="Dark Pink" RGB="0.91 0.58 0.75" />

    <ColorCycle name="Light Orange" RGB="1.00 0.72 0.43" />

    <ColorCycle name="Aquamarine" RGB="0.42 0.99 0.87" />

    <ColorCycle name="Olive Green" RGB="0.85 0.87 0.35" />

    <ColorCycle name="Brown" RGB="0.91 0.46 0.31" />

    <ColorCycle name="Light Blue" RGB="0.33 0.85 0.97" />

    <ColorCycle name="Bright Pink" RGB="0.91 0.31 0.65" />

    <ColorCycle name="Pink" RGB="0.99 0.58 0.81" />

    <ColorCycle name="Pastel Pink" RGB="0.91 0.63 0.58" />

    <ColorCycle name="Gold" RGB="0.91 0.81 0.39" />

    <ColorCycle name="Cerulean Blue" RGB="0.38 0.62 0.89" />

    <ColorCycle name="Dark Green" RGB="0.52 0.87 0.35" />

    <ColorCycle name="Light Brown" RGB="0.91 0.62 0.31" />

    <ColorCycle name="Royal Blue" RGB="0.28 0.33 0.84" />

    <ColorCycle name="Red" RGB="0.91 0.36 0.31" />

    <ColorCycle name="Dark Pink" RGB="0.95 0.53 0.58" />

    <ColorCycle name="Peach Orange" RGB="1.00 0.72 0.62" />

    <ColorCycle name="Yellow" RGB="1.00 0.95 0.41" />

    <ColorCycle name="Light Sky Blue" RGB="0.47 0.84 0.98" />

    <ColorCycle name="Light Green" RGB="0.38 0.98 0.38" />

    <ColorCycle name="Bright Orange" RGB="1.00 0.75 0.33" />

    <ColorCycle name="Indigo Blue" RGB="0.42 0.31 0.97" />

    <ColorCycle name="Coral Orange" RGB="1.00 0.48 0.32" />

  </ColorCycleTable>

Collapse

This table is stored in the install directory for Fusion.  There is a file for each color scheme.  For instance, if you use Photo Booth (the default), it is PhotoBooth.xml.  You are more than welcome to hunt down this file and modify it to your heart's content.  But, it will get overwritten on updates, I think, so this may become tedious.

One more small point.  You said: "I think you may be seeing the current version of that file. This version reflects the fact that I went back after learning about Cycle Component Color and manually fixed the colors to make them more distinct".  That is not possible.  The cycling colors are stored on the machine as part of the install, they are not stored with the design, so you cannot have modified the colors on my installation.  I think, though, that the process by which colors are assigned is somewhat arbitrary, and I may have just lucked out and got a more distinct set of colors.  I can say with some certainty that, until you have 37 components in your design (37 unique components, not counting multiple instances of a component) there should never be any exact duplicates.  If there truly are duplicates, then that is a bug in the logic that assigns the cycling colors.

[GRSnyder]

Thanks Jeff, this is great information. This could lead to an infinite tunnel of tweaking, but I don't think I can resist trying out some other options. 🙂 And as always, I really appreciate your and the other Autodesk folks' willingness to engage on these forums.

---

@jeff_strater wrote: The cycling colors are stored on the machine as part of the install, they are not stored with the design, so you cannot have modified the colors on my installation.  I think, though, that the process by which colors are assigned is somewhat arbitrary, and I may have just lucked out and got a more distinct set of colors.

---

You do seem to be seeing the same colors as me. Below, you on the left, me on the right. I updated the color for Blower after you pointed out the similarity.

[lichtzeichenanlage]

Same on my side. I have to call "Cycle Component Color" command much more often than 6 month ago. Sometimes even with just 3 or 4 components in the design. 

[lichtzeichenanlage]

Here is another example / issue with the color cycle. I can distinguish the colors in the browser but not in the model itself. 

and the reason is: