Keyword
  • Page:
  • 1
  • 2

TOPIC:

Visualizing Tamiya Paint Colors 5 years 4 months ago #52873

OK, I've roughed out a basic color similarity tool .



The closest paint is shown for each of the five categories (X, XF, PS, PC, TS). Within each group, self-similarity is excluded, but it's allowed between groups -- so for example PS2 Red matches PC2 Red.

Worth noting is that this work is still being done on the hex color values of the color swatches Tamiya has on their site. These are wildly inaccurate but good enough to code a proof-of-concept. :)

How do we determine color "similarity"? This is a tough problem to solve because colors that are the different hues are seen by us perceptually to have different lightness, even though they may evaluate identically on that scale colorimetrically. The way to solve this is with ΔE (Delta E) , an equation that accounts for these perceptual variations. What I've built uses an implementation of Delta E 2000, a fun little formula that goes something like this...



:blink: Anyway, I haven't tested this extensively but wanted to post it here to see what folks thought: link . It will get a lot better once I measure the actual paints with a colorimeter.
Attachments:

Please Log in to join the conversation.

Last edit: by pleonard.

Visualizing Tamiya Paint Colors 5 years 4 months ago #52874

I love it!
The functionality is great, but maybe offer some way for the user to copy/paste (or even email) the results from the comparison?
You should also, when time comes, add info on how the comparison is done (layman's terms) and how the colours has been measured, to increase confidence for the user.

One issue, the "tooltip" box gets very crammed/unreadable on colours close to the right border of the screen.
Signature now gets correct formatting if you edit it in your Profile. Use normal BBcode if you want.
www.tamiyabase.com was formatted automatically.

Please Log in to join the conversation.

Last edit: by larbut.

Visualizing Tamiya Paint Colors 5 years 4 months ago #52879

Some bugfixes:

1) Fixed handling of "PS" paints without numbers (limited editions)
2) Removed TS13 Clear
3) Tooltip now shows all caps paint names
4) Switched tooltip from opacity:1 to display:inline for performance reasons
5) Tooltip now tracks mouse smoothly
6) Introduced x buffer to avoid tooltips going off right side of page
The following user(s) Liked this: larbut

Please Log in to join the conversation.

Visualizing Tamiya Paint Colors 5 years 4 months ago #52880

Again, I applaud you for you efforts, and (using the Dogbert scale of bug severity) for avoiding "lethal" and "boneheaded" issues :y: ... however there are a couple that are at least "vexing":

- bottom two rows of the PS table are pretty much unusable in Firefox as the tooltip extends off the bottom of the screen, if you scroll down the rest is very briefly visible before it pings back;
- I know you said the calculated similar colours would improve with measurement of actual painted results, however I have to question whether that's going to work where no visually similar (to my eye) exists. For instance, looking at the tooltip for X17 pink, there is no further similar X or XF shade, but X11 Chrome Silver and XF80 Royal Light grey come up. Under no circumstances I can imagine would someone choose to use those as alternatives, at best you might mix white with a tiny bit of red for a gloss finish, for flat you'd mix X17 with some flat base. I suspect it comes down to what you want the tool to do: 1) provide valid alternatives to modelmakers (in which case sift out the unrealistic similarities manually); or 2) prove your maths works (in which case filter out results like those where the "nest nearest" result is beyond a certain bound).

:)

Please Log in to join the conversation.

Visualizing Tamiya Paint Colors 5 years 4 months ago #52914

All good points! I'm playing around with showing the similars at a constant place at the top of the screen:

thegogglesdonot...acolor/v3/

Still needs to be done:

1) Provide a way to copy or otherwise export the results (Lars förslag)
2) Provide more clarity on how the recommendations are produced (Jonny's suggestion)

Jonny is absolutely right that some of the results make sense and some don't. This is because it's looking for the closest color within each class of paints separately. In some cases -- especially those with 70 colors to choose from -- the nearest one is quite close. And in others -- say, trying to match Pink within XF flats -- the nearest color is nowhere close.

The suggestion of thresholding is a good one, as is visually representing how close each match is. (In such a system, the XF grey that nominally matches pink would be visibly further away from the other matches, which are actually pink. What I'll have to do to accomplish that is to store the relative closeness result and display it. (Currently I use an accumulator that just keeps the closest match within each group, discarding the others as well as all of the similarity measurements.)

Please Log in to join the conversation.

Last edit: by pleonard.
  • Page:
  • 1
  • 2
Moderators: caprinutstingray-63AndyAus
Time to create page: 0.295 seconds
Cookies are required to make this site work. If you continue to use this site you permit us to use cookies.