cluster_analyzer.py

import numpy as np
from sklearn.cluster import KMeans
from sklearn.decomposition import PCA
from sklearn.preprocessing import StandardScaler
import matplotlib.pyplot as plt
import seaborn as sns
from config import NUM_CLUSTERS, OUTPUT_DIR
import os

class ClusterAnalyzer:
    def __init__(self, n_clusters=NUM_CLUSTERS):
        self.n_clusters = n_clusters
        self.scaler = StandardScaler()
        self.kmeans = None
        self.pca = None

    def fit_predict(self, features):
        """Fit KMeans and return cluster labels"""
        # Standardize features
        features_scaled = self.scaler.fit_transform(features)

        # Adaptive PCA - use min(n_samples, n_features, 50) components
        n_components = min(features_scaled.shape[0] - 1, features_scaled.shape[1], 50)
        if n_components < 1:
            n_components = 1

        print(f"Using {n_components} PCA components (adapted to data size)")
        self.pca = PCA(n_components=n_components)
        features_reduced = self.pca.fit_transform(features_scaled)

        # Adjust number of clusters if needed
        n_clusters = min(self.n_clusters, len(features_reduced))
        if n_clusters < 1:
            n_clusters = 1

        print(f"Using {n_clusters} clusters")
        self.kmeans = KMeans(n_clusters=n_clusters, random_state=42, n_init=10)
        labels = self.kmeans.fit_predict(features_reduced)
        return labels, features_reduced

    def get_cluster_centers(self):
        """Return cluster centers"""
        if self.kmeans is not None:
            return self.kmeans.cluster_centers_
        return None

    def visualize_clusters(self, features, labels, image_paths, save_path=None):
        """Visualize clusters using PCA"""
        # Further reduce to 2D for visualization (if possible)
        if features.shape[0] > 2 and features.shape[1] > 2:
            pca_2d = PCA(n_components=min(2, features.shape[0] - 1, features.shape[1]))
            features_2d = pca_2d.fit_transform(features)
        else:
            # If we can't do PCA, use first 2 features
            features_2d = features[:, :2] if features.shape[1] >= 2 else np.hstack([features, np.zeros((features.shape[0], 2 - features.shape[1]))])

        # Create plot
        plt.figure(figsize=(12, 8))

        # Handle case where we have only one cluster
        unique_labels = np.unique(labels)
        if len(unique_labels) > 1:
            scatter = plt.scatter(features_2d[:, 0], features_2d[:, 1], c=labels, cmap='tab10', alpha=0.7, s=100)
            plt.colorbar(scatter)
        else:
            plt.scatter(features_2d[:, 0], features_2d[:, 1], c='blue', alpha=0.7, s=100)
            plt.title(f'All samples in single cluster (Cluster {labels[0]})')

        plt.title('Drill Core Sample Clusters (PCA Visualization)', fontsize=16)
        plt.xlabel('Feature Dimension 1')
        plt.ylabel('Feature Dimension 2')

        # Annotate some points
        for i in range(min(15, len(features_2d))):
            if i < len(image_paths):
                filename = os.path.basename(image_paths[i])[:15] + "..."
                plt.annotate(filename, (features_2d[i, 0], features_2d[i, 1]),
                            xytext=(5, 5), textcoords='offset points', fontsize=8, alpha=0.7)

        plt.tight_layout()
        if save_path:
            plt.savefig(save_path, dpi=300, bbox_inches='tight')
            print(f"Cluster visualization saved to {save_path}")
        plt.show()

    def create_cluster_map(self, image_paths, labels):
        """Create mapping from cluster ID to image paths"""
        cluster_map = {}
        for path, label in zip(image_paths, labels):
            if label not in cluster_map:
                cluster_map[label] = []
            cluster_map[label].append(path)
        return cluster_map

    def analyze_cluster_characteristics(self, features, labels, image_paths):
        """Analyze characteristics of each cluster"""
        cluster_stats = {}

        # Get features for each cluster
        for cluster_id in np.unique(labels):
            mask = labels == cluster_id
            cluster_features = features[mask]

            # Calculate statistics
            mean_features = np.mean(cluster_features, axis=0)
            std_features = np.std(cluster_features, axis=0)

            # Get image paths for this cluster
            cluster_images = [path for i, path in enumerate(image_paths) if labels[i] == cluster_id]

            cluster_stats[cluster_id] = {
                'count': len(cluster_images),
                'mean_features': mean_features,
                'std_features': std_features,
                'sample_images': cluster_images[:5]  # First 5 samples
            }

        return cluster_stats

    def analyze_clusters(self, features, image_paths):
        """Complete clustering analysis"""
        print(f"Performing clustering analysis on {len(image_paths)} samples...")
        print(f"Feature dimensions: {features.shape}")

        # Perform clustering
        labels, features_reduced = self.fit_predict(features)

        # Create cluster map
        cluster_map = self.create_cluster_map(image_paths, labels)

        # Analyze cluster characteristics
        cluster_stats = self.analyze_cluster_characteristics(features, labels, image_paths)

        # Visualize if we have enough samples
        if len(image_paths) > 2:
            viz_path = os.path.join(OUTPUT_DIR, "clusters.png")
            self.visualize_clusters(features, labels, image_paths, viz_path)

        # Print cluster information
        print("\n" + "="*60)
        print("CLUSTER ANALYSIS RESULTS")
        print("="*60)
        for cluster_id, stats in cluster_stats.items():
            print(f"\nCluster {cluster_id}:")
            print(f"  Samples: {stats['count']} images")
            print(f"  Sample files:")
            for path in stats['sample_images']:
                print(f"    - {os.path.basename(path)}")

        return labels, cluster_map, cluster_stats

if __name__ == "__main__":
    pass