Multi-modal Attention System

Hello and good morning, I am starting this topic because I am currently working on an attention system that takes various methods to determine what to look at, it is called a bottom-up attention system (you can google it).  After which we can pull as much info from those frames allowing us to add in short-term memory long-term memory, and eventually action memory what I am trying to do is give imoov a cognitive mind I am asking for all the help that I can get from fellow programmers as I don't have a physical robot yet.  Don't have a 3d printer (anyone willing to print me a complete head with the modded neck the tilts side to side I would be willing to pay)  I have a lot of the code written mostly in c++ slowly I have some friends who are better at java helping to integrate it into myrobotlab as I know c++ and not java I could use all the help I can get and any ideas would be greatly appreciated.  Let me know what you think

GroG's picture

Hi Mike, Sounds really

Hi Mike,
Sounds really exciting.  Do you have a github account where the c++ is and/or a fork of myrobotlab which you and your friends are currently working ?  If not I would heartily recommend posting your project there and forking mrl.'s picture


Trying to get everything together for you GroG took your advice the other day just takes time getting all this code together's picture

Have you thought about

Have you thought about maybe a ROS or YARP plug-in it would allow C++ code to be used with yours YARP even supports windows and Linux and they make a java library

GroG's picture

Ya .. A ROS bridge has been

Ya .. A ROS bridge has been on the board for a while ... but no one has as yet buckled down and built it.  
With the next release "Nixie" - the focus will be 

  • new build system (maven)
  • new dependency management system (ivy + ivy files)
  • new network overlay

The network overlay will define a lot more regarding inter-connectivity between MRL & any "any" system .. including ROS.   So you can run ROS, ORCAS, YARP or any robototic framework and if it can do simple messaging over the network, then all of MRL will be accessable both for control & feedback.. (similar to ROS pub/sub)'s picture


thanks, that will really cut down on the recoding on parts and the ability to use other library types