Second time posting this. Thanks to everyone who mailed us to report problems. The clustering is fixed and works extremely well now! There's a helpful error message when the firewall at work or on public wifi is dropping Melondream's UDP packets. (Melondream communicates via UDP fountain codes.)
We're looking for someone who's interested in commercializing this technology. The user experience is unprecedented. Right?
Watch random images ("WANDER"). When you like one, see the rest of the photoset on repeat ("TRANCE"). To get more of some detail-- a pose, a hairstyle, some scenery-- press that part of the image ("DREAM").
For example, if you see a woman in panties, press the panties to see more girls wearing panties. If you like a woman's face, press her face to see similar faces. If you like a smile, press it to see more. If want more beach scenes, press the sand and water. If you want to see more girls in the woods, press the greenery. If you like to see girls posing on wooden chairs, press the chair. And so on.
This is a solution to the problem of large-scale unsupervised image clustering.
Tried to install on android emulator, works deadly slow, i can't understand what's the benefit as porn consumer for me. Make webapp, running fast enough, so i at least can use it...
Melondream is an app written in golang (using gomobile). It's almost entirely native code, with a small amount of java Android framework on top. Not clear how to translate such a thing into a webapp.
Surely you can imagine applications beyond pornography! The "soft1" server has tens of thousands of photos, indexed by content. Each photo has a web link (the VISIT button).
As for erotica, you can for example search this huge catalog of nudes by genital (vulva) structure. Find one you like, then see who it's attached to. Yes it's depraved, but...
Or faces. Search for faces with a feature you like.
RichardHeart, as we discussed with Qyron below, the APK has been modified to print golang's standard library net package error message directly to the user.
If you get a moment, could you run the new APK and tell us what the net package is saying?
We expect something like "write to UDP6 ... network unreachable", but if would be informative if the error is something different.
Richard, we reverted this change after Qyron ran his test. Basically Melondream is failing to send an IPv6 UDP packet with an 8-byte payload. Maybe this is an Android 7.0 problem, or a problem with the interaction between golang's net package and Android. Sorry
On the first blank screen I try to input "soft1", then I see some sort of connection dialog but then I get error message which, again, is not visible. Attaching link to Dropbox folder with screenshots. https://www.dropbox.com/sh/tagc4veaeco8umm/AAA2xcvvLQoM_lUh1...
The error message (which you can't see) is probably telling you that your Internet connection is blocking Melondream's UDP packets. Are you connecting from work or public WiFi?
Btw, the lesson here for developers is: don't use UDP, use TCP. We made a big mistake
The system was built from the ground up to use forward error correction over UDP. The point was to allow a much less expensive, super-efficient server. We had no idea how poorly UDP is treated on the Internet. Huge, fatal mistake
Now interface works well. But still no connection. I've tried both cellular internet (where UDP is most probably blocked) and home WiFi which doesn't block UDP (for sure, I tried pinging from phone).
We appreciate any bug reports. If Melondream is failing for you, a comment posted here with your Android version (e.g., Marshmallow 6.0.1) and the failure you're seeing (error message) is valuable to us. Thank you.
We're looking for someone who's interested in commercializing this technology. The user experience is unprecedented. Right?
Melondream http://driftwheeler.com/
Use server ID "soft1"
Watch random images ("WANDER"). When you like one, see the rest of the photoset on repeat ("TRANCE"). To get more of some detail-- a pose, a hairstyle, some scenery-- press that part of the image ("DREAM").
For example, if you see a woman in panties, press the panties to see more girls wearing panties. If you like a woman's face, press her face to see similar faces. If you like a smile, press it to see more. If want more beach scenes, press the sand and water. If you want to see more girls in the woods, press the greenery. If you like to see girls posing on wooden chairs, press the chair. And so on.
This is a solution to the problem of large-scale unsupervised image clustering.
Enjoy!