ascota_core¶
ascota_core is the first stage of the ASCOTA pipeline.
It provides the fundamental computer vision and analysis utilities that later
modules build upon.
Purpose¶
This package contains tools for:
-
Segmentation & detection
Identifying and isolating key elements in input images (e.g., sherds, measurement cards).
Uses models like SAM (Segment Anything Model) together with OpenCV-based preprocessing. -
Color card classification & correction
Detecting reference color cards, classifying patches, and applying color correction to normalize input images across lighting conditions. -
Geometric scaling & surface estimation
Leveraging detected measurement cards to compute pixels-per-centimeter ratios and estimate the surface area of sherds/pottery fragments.
Role in the pipeline¶
ascota_core is the core foundational package that prepares and standardizes
input images for downstream classification and analysis. It ensures that:
- Sherds are properly segmented from the background.
- Color profiles are normalized to a standard reference.
- Geometric measurements are calibrated to real-world units (via scale cards).
Together, these ensure later classification is performed on calibrated, comparable data.