in Iconfinder News

Iconfinder Hackathon – Day 3 and 4

Mid-week, mid-project update. Everyone is getting into the nitty gritty and tackling bugs, design issues and other hurdles. In that way the hackathon is much like any other project form. Once the initial excitement settles, it’s grind time.


Martin – Icon Designer Analytics

After some fiddling with Google Charts, the choice for chart plugin from Amcharts. Most of the work for the past days have been working on the view and JavaScript. So far the analytics page looks like this:

Image 2014-02-28 at 10.48.34 am (1)

Nick – Entangle, a MessagePack based IDL for a service oriented architecture

Despite a bit of biological issues (damn you, fragile human body!) progress on Entangle has been good. The parser is now fully functional (have a look at if you’re interested in the source code) and can interpret definitions into a useful data structure.

Screen Shot 2014-02-26 at 00.08.05

The natural next step is to turn the definition into useful code for target languages. Today has been spent implementing the wire protocol for Entangle in Go, intertwined with research into how others do it to avoid reinventing the entire circumference of the wheel. If all goes well, tomorrow should see Entangle output Go code and the first service fired up — fingers crossed!

— Nick

Silviu – Duplicate Icon Detection

The library can detect identical icons based on hashes.

The next step would be to compare icons by comparing the hamming distance between two average hashes, the hamming distance withing two strings is the number of substitutions required to change a string into the other. A distance of zero indicates that the icons are identical, a distance of 5 means there are only a few things different but they are pretty close to being identical, a distance of 15 or more means that the icons are very different.

Two identical icons with their hash:

Two identical icons with their hash:

Slightly different icons:
Slightly different icons:

Completly different icons:
Completly different icons: >>> IconHash.objects.hamming_distance('c2842973e66dd03f', '0402f0f8b4e1cf1f') 15

— Silviu

Jens – Iconfinder Chrome Extension

My extension now has a working API lookup and shoes matching icon results from iconfinder as you type. You also have the ability to choose a specific style to search for. Next step will be to make an auto selector for the styles. By look for words matching any of the style names, it will be possible to select the maching  style and thereby search on both the tearm and typed style without touching the mouse.

Screen Shot 2014-02-27 at 17.31.19

— Jens

Jess – Referral Program

After deciding on a two-way incentive and designing a basic referral page, I went on to write some sample tweets and Facebook updates. I also designed a basic email, and am now working on the flow for onboarding an invited user. This is the detailed part of the project. Trying to design a really smooth and delightful first time experience for someone who has never been to Iconfinder before.


Second thing that also requires a lot of details work is fraud prevention. This is a very complex issue since it can have great influence of the success of the program. And in more than one way. If the program is to simple, it could be easily gamed, which would lead to bad economics for the program. But if the program is to complex and requires to much of the user, it would be more fraud safe, but might also limit the promotion of the program. It’s a delicate balance.

— Jess

Write a Comment