Skip to content
No description, website, or topics provided.
HTML Python Makefile
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
figures
2019-07-15_CNS.pdf
2019-07-15_CNS.py
LICENSE
Makefile
README.md
index.html

README.md

Learning where to look: a foveated visuomotor control model

  • What:: talk @ conference

  • Who:: Emmanuel Daucé, Pierre Albigès & Laurent Perrinet

  • Where: Barcelona (Spain), see https://laurentperrinet.github.io/talk/2019-07-15-cns

  • When: 15/07/2019, time: 12:20:00-12:40:00

  • What:

    • Slides @ https://SpikeAI.github.io/2019-07-15_CNS
    • Code for slides @ https://github.com/SpikeAI/2019-07-15_CNS/
    • Abstract: In computer vision, the visual search task consists in extracting a scarce and specific visual information (the ``target'') from a large and crowded visual display. This task is usually implemented by scanning the different possible target identities at all possible spatial positions, hence with strong computational load. The human visual system employs a different strategy, combining a foveated sensor with the capacity to rapidly move the center of fixation using saccades. Saccade-based visual exploration can be idealized as an inference process, assuming that the target position and category are independently drawn from a common generative process. Knowing that process, visual processing is then separated in two specialized pathways, the "where" pathway mainly conveying information about target position in peripheral space, and the "what" pathway mainly conveying information about the category of the target. We consider here a dual neural network architecture learning independently where to look and then at what to see. This allows in particular to infer target position in retinotopic coordinates, independently to its category. This framework was tested on a simple task of finding digits in a large, cluttered image. Simulation results demonstrate the benefit of specifically learning where to look before actually knowing the target category. The approach is also energy-efficient as it includes the strong compression rate performed at the sensor level, by retina and V1 encoding, which is preserved up to the action selection level, highlighting the advantages of bio-mimetic strategies with regards to traditional computer vision when computing resources are at stake.
You can’t perform that action at this time.