Jump to content

  • Log In with Google      Sign In   
  • Create Account

Interested in a FREE copy of HTML5 game maker Construct 2?

We'll be giving away three Personal Edition licences in next Tuesday's GDNet Direct email newsletter!

Sign up from the right-hand sidebar on our homepage and read Tuesday's newsletter for details!


We're also offering banner ads on our site from just $5! 1. Details HERE. 2. GDNet+ Subscriptions HERE. 3. Ad upload HERE.


#Actualrouncer

Posted 20 April 2013 - 12:16 PM

Im developing this kinda wierd new way to control video with a joystick.
 
ok so say youve got 90 minutes of footage, which has to have "similarities" (remember similarity is just as epic a word as singularity, just we use
it like it isnt special.) all through it, it sorta needs to be from a single camera placement, and the action sorta is "blendable".
 
now i want you to go frame by frame and mark "whats happening" into say 9 different categories, which can also
be on at the same time.
Like for instance "steer left" "steer right" "shoot gun" etc.
 
now.  with some simple system, itll sorta flick through the movie whenever you want one or more of the states, with a spacial match. (say you select categories with
a joystick?)
 
now to get it perfect.   you use this system abbreviated HTM (hierarchical temporal memory), and it makes the following things possible right out of the naive implementation.
 
you can have multiple frames playing at the same time in different parts of the screen.
 
and you have a balance of how "flicky" the changes are, and how non interactive it is,  the more "flicky" it is, the more interactive, the
less flicky the less interactive.
the more you "pool" (destroy/shrink data between regions) the more itll find similarities in differences.
 
a mud basher game with a sprint car, around a tube track would be good, for example.  (note there shouldnt be any moving objects, like no other cars in the race, theres limitations, but a crowd would be ok.)
 
you could just pick the most common frame on the screen and make that have the audio responsibility.
 
would be pretty wacky.

#1rouncer

Posted 20 April 2013 - 12:13 PM

Im developing this kinda wierd new way to control video with a joystick.
 
ok so say youve got 90 minutes of footage, which has to have "similarities" (remember similarity is just as epic a word as singularity, just we use
it like it isnt special.) all through it, it sorta needs to be from a single camera placement, and the action sorta is "blendable".
 
now i want you to go frame by frame and mark "whats happening" into say 9 different categories, which can also
be on at the same time.
Like for instance "steer left" "steer right" "shoot gun" etc.
 
now.  with some simple system, itll sorta flick through the movie whenever you want one or more of the states. (say you select categories with
a joystick?)
 
now to get it perfect.   you use this system abbreviated HTM (hierarchical temporal memory), and it makes the following things possible right out of the naive implementation.
 
you can have multiple frames playing at the same time in different parts of the screen.
 
and you have a balance of how "flicky" the changes are, and how non interactive it is,  the more "flicky" it is, the more interactive, the
less flicky the less interactive.
the more you "pool" (destroy/shrink data between regions) the more itll find similarities in differences.
 
a mud basher game with a sprint car, around a tube track would be good, for example.  (note there shouldnt be any moving objects, like no other cars in the race, theres limitations, but a crowd would be ok.)
 
you could just pick the most common frame on the screen and make that have the audio responsibility.
 
would be pretty wacky.

PARTNERS