analyze_certainty.py

import os
import sys
import numpy as np
import librosa
from tensorflow.keras.models import load_model

# Import project utils
sys.path.append(os.getcwd())
from utils.hear_extractor import HeARExtractor
from utils.audio_preprocessor import advanced_preprocess

# Config
MODEL_PATH = r"c:\Users\ASUS\lung_ai_project\models\hear_classifier_advanced.h5"
CLASSES_PATH = r"c:\Users\ASUS\lung_ai_project\models\hear_classes_advanced.npy"

files = [
    r"C:\Users\ASUS\Downloads\WhatsApp Audio 2026-02-20 at 1.46.51 PM.mpeg", # Correct Healthy (79%)
    r"C:\Users\ASUS\Downloads\WhatsApp Audio 2026-02-20 at 1.52.19 PM.mpeg", # Correct Healthy (67%)
    r"C:\Users\ASUS\Downloads\WhatsApp Audio 2026-02-20 at 2.06.03 PM.mpeg"  # Misclassified Healthy (52% Sick)
]

def analyze_certainty():
    extractor = HeARExtractor()
    model = load_model(MODEL_PATH, compile=False)
    classes = np.load(CLASSES_PATH)
    
    print(f"{'File Name':<35} | {'Pred':<8} | {'Prob Healthy':<13} | {'Prob Sick':<10}")
    print("-" * 75)
    
    for f in files:
        if not os.path.exists(f): 
            print(f"File {f} not found")
            continue
            
        y, sr = librosa.load(f, sr=16000, duration=5.0)
        y_clean = advanced_preprocess(y, sr)
        emb = extractor.extract(y_clean)
        
        if emb is not None:
            probs = model.predict(emb[np.newaxis, ...], verbose=0)[0]
            # Assumes classes are ['healthy', 'sick']
            h_prob = probs[0] if classes[0] == 'healthy' else probs[1]
            s_prob = probs[1] if classes[1] == 'sick' else probs[0]
            pred = classes[np.argmax(probs)]
            
            print(f"{os.path.basename(f):<35} | {pred:<8} | {h_prob*100:>11.2f}% | {s_prob*100:>8.2f}%")

if __name__ == "__main__":
    analyze_certainty()