Robo Tagging Flickr

Yesterday I finished a script to tag photos in flickr with an average RGB hex value. It was created so that a student would be able to seach the Color Fields group on flickr for images based on color. I started the script yesterday at 4:30pm and it was done around 6:30 I am guessing. Around 4:45ish emails start to come in from people curious as to what the hex code meant. Overall the responses have been positive with only one person getting a slight twist in their undies. Surprisingly, this little experiment has gotten me 1,300 views,a bunch of favorites and comments in less then 24hrs. Spamming does work!

Other interesting notes: Flickr’s api authentication system is overly complicated.

Update: Here is the first quick stab at a color viewer. Be patient as it is loading lots of images. Rollover the color to see the thumbnail and click to get to the full image. Here it is

13 Responses to “Robo Tagging Flickr”

  1. kari Says:

    So that was what the tags were about! Cool… maybe I can learn from this and know what to do better with my camera.. interesting..

  2. David Says:

    Hrm. I was going to ask what the deal was with the hexidecimal tag you left on my pic. Now I know.


    All of the tagged pictures will be in the color fields group? or how will it work?

  3. Brettf Says:

    The tagged photos are in the color fields and monocolor groups. Not entirely sure how they will work yet. Likely any searches will have to be a range of color tags.

  4. Eamonn Says:

    One thing I noticed is that each of the tags is effectively unique. It is highly unlikely that more than one photo will have the same RGB value given that there are more than 16 million possible values.

    It is not possible to search a range of tag values in any efficient manner unless you read all tags and cache them locally first.

    So maybe instead of tagging with 24-bit values like 0x84501c you could tag with 12 bit values such as: 0x851 where you only use the 4 most significant bits each of red, green, or blue channel. That reduces the number of possible values to about 4000, so you are more likely to find photos with the same tag. Even that is more colours than most people can distinguish, so probably it is better to use only three bits per channel to get 512 possible values or even 2 bits per channel to get 64 different values.

    Even better would be to convert from RGB to LAB colorspace because it is designed so that equal distances in LAB space correspond to to equal perceptual distance in the human visual system. That means that each of the cells into which you have split up the colorspace will be of equal size to a human.

    Or another approach would be to tag each channel separately using three tags such as red=0x84 green=0x50 blue=0x1c or red=0x8 green=0x5 blue=0x1. Or in LAB space L=… A=… B=…

    I look forward to your future blog posts about this research.

  5. Brettf Says:

    Yeah i thought about most of this. But since i will be searching from and application i can create a perceptual range of colors to search and send a bunch of tags to flickr in the search query. It should work out. Likely I will be converting to LAB or something similar on the client side then creating the tag range and sending that to flickr. Also the RGB space will make it easier to do some apps much faster. Stay tuned.

  6. radiant guy Says:

    I was going to PM you at flickr to thank you for the weird tags LOL
    When I saw them I looked exactly like that ~:O~
    Usually I checked people’s flickr profiles and I found out that you are a teacher so I thought that it’s could be something related to school or something and that’s when I noticed your blog’s link there and that’s why I am here.

    Usually every single activity attracts me to check people’s profile even when they add a photo to their favorite list or when they blog a photo, many other people are acts the way I do so you shouldn’t be surprised when you get more traffic.

    You could find some people similar to “one person” you mentioned don’t let them get into you, you didn’t do any harm.

  7. Bryce Says:

    Yeah, I’m one of the ones who found this site because of the tags. I have no problem w/ the concept, but my big question is what do your students plan on doing w/ the photos they search out, and do they plan to respect the licenses that a photographer may have in place? If these students are going to be using the photos in projects, etc., I certainly hope that they do respect them. Otherwise, I can see where more than the “one person” is going to get a bit bent out of shape.

  8. Brettf Says:

    The school I teach at takes copyright very seriously so I can assure you they will not be used inappropriately.

  9. fiona j Says:

    i was suspicious when i first spotted the unsolicited tags, so checked you out. effective, indeed!

    having read all this i respect the idea of searching for images of a given colour… but what are your students likely to do with the images once found?

  10. Brettf Says:

    I am not entirely sure really. The first set of ideas is to create a mood board generator for a given color palette.

  11. Todd Smith Says:

    Congrats sir, I’m glad that this project is finally “off the ground,” if we can call it that, I remember the first elevator pitch you gave me on this almost a year ago now.

    In response to the peoples concerns about legality and uses, I don’t know the depth of the code, but is it already searching strictly creative commons images? Seems like a step that you would have went through, but I don’t know your code.

    If concerned photographers are reading this, feel secure in Brett’s respect for copyright, I helped him enforce it among students when working side-by-side with him and can attest to it being by-the-book.

    Expect ethical, credited uses. Good Work Brett, I look forward to the next stage.

  12. jan Says:

    Like the others, I wound up here because of the odd tags added to some of my pictures.

    Only problem now is I’m worried that my photos will be used as examples of “what not to do”.

  13. hoopyfrood Says:

    Some crazy cool shit going on here, Forsyth.

    Do you know where I can find some Flash-based software for the refrigerated shipping industry?