Untitled

import numpy as np
import tensorflow as tf

def calculate_metrics(predictions, labels, threshold=0.5):
    # Threshold predictions
    predictions_binary = (predictions > threshold).astype(np.int32)

    # Calculate correct and incorrect predictions
    correct_predictions = np.sum(predictions_binary == labels)
    incorrect_predictions = np.sum(predictions_binary != labels)

    # Calculate misses (false negatives)
    misses = np.sum((predictions_binary == 0) & (labels == 1))

    # Calculate IoU per segment
    intersection = np.logical_and(predictions_binary, labels)
    union = np.logical_or(predictions_binary, labels)
    iou = np.sum(intersection) / np.sum(union)

    return correct_predictions, incorrect_predictions, misses, iou

# Example usage:
# Assuming `model_output` is your model's output, and `true_labels` are your true labels
model_output = ...  # Your model's output
true_labels = ...  # Your ground truth labels

correct, incorrect, misses, iou = calculate_metrics(model_output, true_labels)

print(f"Correct Predictions: {correct}")
print(f"Incorrect Predictions: {incorrect}")
print(f"Misses: {misses}")
print(f"IoU: {iou}")