src/metrics.py

import numpy as np

def calculate_iou(pred_mask, true_mask):
    """
    Calculate Intersection over Union (IoU) for binary masks.
    """
    intersection = np.logical_and(pred_mask, true_mask).sum()
    union = np.logical_or(pred_mask, true_mask).sum()
    if union == 0:
        return 1.0 # If both are empty, perfect match
    return intersection / union

def calculate_dice(pred_mask, true_mask):
    """
    Calculate Dice Coefficient for binary masks.
    """
    intersection = np.logical_and(pred_mask, true_mask).sum()
    return (2. * intersection) / (pred_mask.sum() + true_mask.sum() + 1e-6)

def calculate_accuracy(preds, labels):
    """
    Calculate simple classification accuracy.
    """
    correct = (preds == labels).sum().item()
    total = labels.size(0)
    return correct / total

def calculate_ndvi(nir_band, red_band):
    """
    Calculate NDVI given NIR and Red bands.
    NDVI = (NIR - Red) / (NIR + Red)
    """
    nir = nir_band.astype(float)
    red = red_band.astype(float)
    
    denominator = (nir + red)
    # Avoid division by zero
    denominator[denominator == 0] = 1e-6
    
    ndvi = (nir - red) / denominator
    return ndvi

def ndvi_correlation(predicted_greenery_percentages, true_mean_ndvis):
    """
    Calculate Pearson correlation coefficient between predicted greenery 
    percentage and true mean NDVI across a set of images.
    """
    if len(predicted_greenery_percentages) < 2:
        return 0.0
    return np.corrcoef(predicted_greenery_percentages, true_mean_ndvis)[0, 1]