Skip to content

batch size of 32 do not generate useful segmentation models (64 does) #105

Open
@dcianferoni

Description

@dcianferoni

For some reason when I train using batch size of 32 the accuracy is below 0.5 and visualizing prediction leds to all points predicted with same class. Increasing to 64 leads to good performances. Any explanation?

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions