app.py

# AUTOGENERATED! DO NOT EDIT! File to edit: drum_sample_categorization.ipynb.

# %% auto 0
__all__ = ['learn', 'iface', 'create_spectogram', 'spectogram_to_image', 'predict']

# %% drum_sample_categorization.ipynb 2
import PIL
import gradio as gr
from fastai.vision.all import *
import librosa
import matplotlib.pyplot as plt
import numpy as np
import shutil
import os

# %% drum_sample_categorization.ipynb 4
learn = load_learner('./model.pkl')

# %% drum_sample_categorization.ipynb 6
def create_spectogram(path):
    samples, sample_rate = librosa.load(path, duration=5.0)
    S = librosa.feature.melspectrogram(y=samples, sr=sample_rate)
    return S

def spectogram_to_image(spec):
    fig = plt.figure(figsize=[0.72, 0.72])
    ax = fig.add_subplot(111)
    ax.axes.get_xaxis().set_visible(False)
    ax.axes.get_yaxis().set_visible(False)
    ax.set_frame_on(False)
    librosa.display.specshow(librosa.power_to_db(spec, ref=np.max))
    fig.canvas.draw()
    return PIL.Image.frombytes('RGB',fig.canvas.get_width_height(),fig.canvas.tostring_rgb())

def predict(audio_file_path):
    spec = create_spectogram(audio_file_path)
    spec_image = spectogram_to_image(spec)
    prediction,other,probs = learn.predict(spec_image)
    return prediction

# %% drum_sample_categorization.ipynb 8
iface = gr.Interface(fn=predict, inputs=[
                     gr.components.Audio(type='filepath')], outputs="text")
iface.launch()