Spaces:
Running
Running
File size: 2,981 Bytes
51c4b3f |
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 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 |
import librosa
import tensorflow as tf
import soundfile
import gradio as gr
import pandas as pd
import os
import random
import numpy as np
# Set the environment variable
os.environ["KMP_DUPLICATE_LIB_OK"] = "TRUE"
audio_files_path = 'examples'
c_names = ['Bronchiectasis', 'COPD', 'Healthy', 'Pneumonia', 'URTI']
# Loading Audio Files
audio_files = []
for file in os.listdir(audio_files_path):
if file.endswith('.wav'):
audio_files.append(file)
# convert the list to a df
audio_files_df = pd.DataFrame(audio_files, columns=['audio_paths'])
# print(audio_files_df.iloc[0]['audio_file'])
audio_files_to_show = audio_files_df['audio_paths'].tolist()
# create a gradio interface
# 0. Load models
# 1. Audio File input
# 2. clear and Submit button
# 3. Upon submit , first preprocess the audio file using log mel and then run the outputs through the AI model
# # 4. Output the prediction
def load_model():
# Load the model
return tf.keras.models.load_model("models/lung_disease_predictor_cnn_logmel_without_data_augmentation.keras")
def preprocessing(audio_file, mode):
# we want to resample audio to 16 kHz
sr_new = 16000 # 16kHz sample rate
x, sr = librosa.load(audio_file, sr=sr_new)
# padding sound
# because duration of sound is dominantly 20 s and all of sample rate is 22050
# we want to pad or truncated sound which is below or above 20 s respectively
max_len = 5 * sr_new # length of sound array = time x sample rate
if x.shape[0] < max_len:
# padding with zero
pad_width = max_len - x.shape[0]
x = np.pad(x, (0, pad_width))
elif x.shape[0] > max_len:
# truncated
x = x[:max_len]
if mode == 'mfcc':
feature = librosa.feature.mfcc(y=x, sr=sr_new)
elif mode == 'log_mel':
feature = librosa.feature.melspectrogram(y=x, sr=sr_new, n_mels=128, fmax=8000)
feature = librosa.power_to_db(feature, ref=np.max)
return feature
def predict_lung_disease(audio_data):
# Create a temporary file
filename = "temp/lungs_audio.wav" # Set your desired filename
soundfile.write(filename, audio_data[1],samplerate=audio_data[0]) # Save audio to file
# Process the temporary audio file
processed_audio = preprocessing(filename, 'log_mel').reshape((-1, 128, 157, 1))
new_preds = model.predict(processed_audio)
new_classpreds = np.argmax(new_preds, axis=1)
print(str(c_names[new_classpreds[0]]))
return str(c_names[new_classpreds[0]])
# Gradio Interface
model = load_model()
# have example audio files to test
# Interface
iface = gr.Interface(
fn=predict_lung_disease,
inputs=["audio"],
outputs="text",
title="VitalSenseAI",
examples=
[
[os.path.join(audio_files_path,audio_file)] for audio_file in audio_files_to_show
]
)
iface.launch()
|