Skip to content

Test benchmark by changing feature extraction segment erasing with 0 values pixels outside of superpixel  #40

@ryukinix

Description

@ryukinix

At first, change default value of erase_color parameter to 0.

egsis/egsis/features.py

Lines 163 to 171 in cc477cf

def feature_extraction_segment(
img: np.ndarray,
segments: np.ndarray,
label: int,
max_radius: Optional[int] = None,
centroid: Optional[List[int]] = None,
erase_color: Optional[int] = None,
feature_method: FeaturesMethods = "comatrix"
):

After running the benchmark, if metrics was improved by this change, consider erase_color being some kind of hyperparameter for EGSIS.

Metadata

Metadata

Assignees

No one assigned

    Labels

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions