ascota_classification

ascota_classification is the second stage of the ASCOTA pipeline. It provides model-based classification and feature clustering utilities for preprocessed pottery imagery.

Purpose

This package focuses on:

• Type classification Multi-stage pottery type classification pipelines, including optional appendage subtype refinement.
• Decoration classification DINOv2 feature extraction with a trained classifier for decoration labels.
• Color clustering Lab-based feature extraction and HDBSCAN clustering for visually similar images.
• Texture clustering Texture feature extraction (LBP + GLCM), PCA projection, and HDBSCAN grouping.

Submodules

• color: Lab feature extraction and color similarity clustering.
• texture: texture feature extraction and clustering.
• decoration: pottery decoration classification.
• type: pottery type and pottery/non-pottery classification.