8000 Make convolutions more efficient · Issue #7 · WinawerLab/sco · GitHub
[go: up one dir, main page]
More Web Proxy on the site http://driver.im/
Skip to content
Make convolutions more efficient #7
Open
@billbrod

Description

@billbrod

Currently, we only produce filters for the spatial frequencies specified by the dictionary returned by the frequency preference function. If your frequencies are very low (e.g., 0.2 or 0.4), the convolutions can take a very long time (several hours, as opposed to several minutes) to run. In order to make this more reasonable, we probably want to do something like shrinking the image first (losing the high frequency signal, but a lowpass filter doesn't care about that anyway) and then applying the filter, so there are fewer points to convolute it with. There are probably efficient implementations of this in python somewhere.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions

      0