Homescreen
p/homescreen
A tool for sharing your home screen and discovering apps
Hot Homescreen Apps — You can now see trending new apps from homescreens.
Featured
13
Replies
Harry Raymond
Good way to identify your app users too.
Andrew Bryk
@matthartman This is awesome! I'd love to be able to click on the tags to be able to sort and filter specific categories of apps (EX: I'm not big into games so would want to filter those out)
Toby Padilla
@matthartman @andrewjbryk the tags you see on the New Apps are currently being pulled from the App Store. In an upcoming rev, we're going to be launching our own categories that hopefully will be much better! We'll include category level charts and navigation as well.
We've been internally testing a @homescreenapp algorithm at @betaworks to identify trending new apps on people's homescreens. We've already uncovered lots of great apps from indie developers so now we're ready to release hot new apps. It also surfaces trending apps early -- while we were testing, we saw Microsoft @Outlook shot up to trend #1, showing up on 0.83% of people's homescreens, indicating people were *actually* using it. We found this to be anecdotally true. More recently, we saw @AppMeerkat trending as soon as @JoshConstine wrote about it in @techcrunch.
Hi @producthunt readers, @toby is talking about this on @_upclose now.
Toby Padilla
For those interested in how this algorithm works: As we identify apps on homescreens, we note which ones are new to the community. Once one of the "new" apps is seen on a few (tweakable threshold) homescreens, it shows up on this new page, chronologically reverse ordered by when it was first discovered. We've found it to be very effective in terms of discovering new apps. Please let us know what you think!
Paolo Perazzo
@toby hi! Very interesting but the main question for me is: "how do you identify apps on people homescreen and I guess their usage?" Thanks!
Toby Padilla
@sivola Hi! We do image recognition the screenshot people upload via our app. We identify the phone model, then slice up the icon grid and compare against out icon db using various image recognition techniques (we're constantly trying new things). We've also been working on some internal tools to help us train our algorithms, correct mistakes and generally make system more accurate.
Taylor Edmiston
@sivola @toby A very cool modern application of image / pattern recognition. It's great to see these techniques at work outside of research labs. Are you able to handle custom font faces, sizes, and colors?
Toby Padilla
@sivola @kicksopenminds not yet so jailbroken phones with extreme customization do break the image recognition. People seem to like publishing them to #Homescreen anyway though, so that's good! :)
Nicholas Sheriff
@toby @matthartman This looks just amazing ugh I'm envious of how delicious this algorithmic approach is....I need this for Quest....let's talk pleaseeee ( do you guys have an api? ) nick@questmobile.co