Closed fredhohman closed 8 years ago
I agree that this could be helpful, especially in movie scenes where the color change can rely on a small bright point. We could have multiple color options to switch between: (1) most frequent colors (2) most contrastive colors (3) background colors, etc.
On Sun, Sep 11, 2016 at 8:56 PM, Fred Hohman notifications@github.com wrote:
Currently get_colors.py gets most frequent pixel colors. Should this change to become the most "dominant" colors? E.g. what if an image was mostly shades of a single color, but had some part that was drastically different that draws a person's eye, however that region's colored pixel count could be small in comparison to the other shades.
Thoughts?
— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/fredhohman/cs-7450/issues/1, or mute the thread https://github.com/notifications/unsubscribe-auth/ABPhM0nrxHdlL6QOV1H9k94nrjP0iI5Hks5qpKM3gaJpZM4J6Jzn .
Sounds like a good idea. Now... any idea how to implement these? Without the frequency count we may lose the ability to rank colors. I suppose most contrastive would be more tractable.
I guess for most contrastive you could get most frequent, then cluster on RGB similarities, then pull out the one or two colors that didn't get assigned a cluster. Not sure how tractable that is for thousands of frames, though.
On Sun, Sep 11, 2016 at 9:09 PM, Fred Hohman notifications@github.com wrote:
Sounds like a good idea. Now... any idea how to implement these? Without the frequency count we may lose the ability to rank colors. I suppose most contrastive would be more tractable.
— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/fredhohman/cs-7450/issues/1#issuecomment-246217820, or mute the thread https://github.com/notifications/unsubscribe-auth/ABPhM6Db3oP7EtKUf-2Ji_Mx16DzWdKSks5qpKY0gaJpZM4J6Jzn .
Actually by removing the sorting lines of code we might get what we want... can try this later.
OK. I think for the proposal we'll just need one kind of color viz but it's good to have other options to implement later.
On Mon, Sep 12, 2016 at 7:28 PM, Fred Hohman notifications@github.com wrote:
Actually by removing the sorting lines of code we might get what we want... can try this later.
— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/fredhohman/cs-7450/issues/1#issuecomment-246527457, or mute the thread https://github.com/notifications/unsubscribe-auth/ABPhMx7PV2TjRCzOuwxUrX6RRMoH9WB9ks5qpeAQgaJpZM4J6Jzn .
Possibly relevant for this.
Currently
get_colors.py
gets most frequent pixel colors. Should this change to become the most "dominant" colors? E.g. what if an image was mostly shades of a single color, but had some part that was drastically different that draws a person's eye, however that region's colored pixel count could be small in comparison to the other shades.Thoughts?