File size: 2,570 Bytes
cfd1552 233adb5 cfd1552 233adb5 cfd1552 233adb5 cfd1552 233adb5 cfd1552 233adb5 cfd1552 233adb5 cfd1552 233adb5 cfd1552 233adb5 cfd1552 233adb5 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 |
import torch
import torchaudio
import librosa
import soundfile as sf
import numpy as np
from src.model.emotion_classifier import EmotionClassifier
from src.model.feature_extractor import feature_extractor, processor
from src.utils.preprocessing import resampler
from src.config import DEVICE, LABELS
import os
# Charger le modèle sauvegardé
classifier = EmotionClassifier(feature_extractor.config.hidden_size, len(LABELS)).to(DEVICE)
classifier.load_state_dict(torch.load(os.path.join("src","model","best_emotion_model.pth"), map_location=torch.device(DEVICE)), strict=False)
classifier.eval()
# Fonction de prédiction
def predict_emotion(speech, output_probs=False, sampling_rate=16000):
# Charger l'audio
# waveform, sample_rate = librosa.load(speech, sr=None)
# speech_audio, sample_rate = sf.read(speech, dtype="float32")
# Rééchantillonnage si nécessaire
# if sample_rate != sampling_rate:
# speech = torch.tensor(speech).unsqueeze(0)
# speech = resampler(speech).squeeze(0).numpy()
# Extraire les features
inputs = processor(speech, sampling_rate=sampling_rate, return_tensors="pt", padding=True)
input_values = inputs.input_values.to(DEVICE)
with torch.no_grad():
features = feature_extractor(input_values).last_hidden_state.mean(dim=1)
logits = classifier(features)
if output_probs:
# Appliquer softmax pour obtenir des probabilités
probabilities = torch.nn.functional.softmax(logits, dim=-1)
# Convertir en numpy array et prendre le premier (et seul) élément
probabilities = probabilities[0].detach().cpu().numpy()
# Créer un dictionnaire associant chaque émotion à sa probabilité
emotion_probabilities = {emotion: prob for emotion, prob in zip(LABELS, probabilities)}
# emotion_probabilities = {"emotions": [emotion for emotion in emotion_labels],
# "probabilities": [prob for prob in probabilities]}
return emotion_probabilities
else:
# Obtenir l'émotion la plus probable (i.e. la prédiction)
predicted_label = torch.argmax(logits, dim=-1).item()
emotion = LABELS[predicted_label]
return emotion
# Exemple d'utilisation
# if __name__ == "__main__":
# base_path = os.path.abspath(os.path.join(os.path.dirname(__file__), "data"))
# audio_file = os.path.join(base_path, "colere", "c1ac.wav")
# emotion = predict_emotion(audio_file)
# print(f"🎤 L'émotion prédite est : {emotion}")
|