-
Notifications
You must be signed in to change notification settings - Fork 0
/
app.py
282 lines (228 loc) · 10.3 KB
/
app.py
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
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
import base64
import os
import zipfile
import tempfile
import pandas as pd
import numpy as np
import streamlit as st
from pydub import AudioSegment
import glob
from IPython.display import Audio
import shutil
import time
import pickle
from tools import get_all_audios_files, load_audio_file
from extract_audio_features import extract_features, extract_features_from_dataset
st.set_page_config(
page_title="Drum Kit Classifier",
page_icon="🥁",
layout="centered",
initial_sidebar_state="expanded",
)
# st.image("logo.jpg")
st.sidebar.title("Drum Kit Classifier 🥁")
st.sidebar.caption("Use machine learning to classify drum kits")
st.sidebar.divider()
st.sidebar.markdown("""
**📝 Instructions:**
1. Upload your drum kit zip file.
2. Select a model.
3. Click on the `Classify` button.
4. Wait for the classification results to appear.
""")
st.sidebar.divider()
st.sidebar.success("This is a **beta** version of the app.")
# =======
# Session state
# =======
# We need to set up session state via st.session_state so that app interactions don't reset the app.
if not "ok_submit" in st.session_state:
st.session_state.ok_submit = False
# =======
# Utils
# =======
# Load your trained models here (example models: model1, model2)
# You may need to adjust this part to suit your actual models
@st.cache_resource(show_spinner=True)
def load_ml_model(model_path):
if not os.path.exists(model_path):
raise FileNotFoundError(f"Model '{model_path}' not found")
items = []
with open(model_path, 'rb') as f:
while True:
try:
items.append(pickle.load(f))
except EOFError:
break
model = items[0]
print(model, "#######################", type(model))
scaler = items[1]
features_columns = items[2]
print(f"> '{model_path}' loaded successfully")
print(f">>> Model: {model}")
print(f">>> Scaler: {scaler}")
print(f">>> {len(features_columns)} Features columns: {features_columns[:5]} ... {features_columns[-5:]}")
return model, scaler, features_columns
RF_MODEL_PATH = './models/RandomForest data[aug=1, s=10291, s_per_class=[596,1283], n_feats=77, feat_select=1] 20230514_04.pkl'
SVC_MODEL_PATH = './models/SVC data[aug=1, s=10291, s_per_class=[596,1283], n_feats=77, feat_select=1] 20230514_15.pkl'
LGBM_MODEL_PATH = './models/LGBM data[aug=1, s=10291, s_per_class=[596,1283], n_feats=77, feat_select=1] 20230515_02.pkl'
model_rf, scaler_rf, features_columns_rf = load_ml_model(RF_MODEL_PATH)
model_svc, scaler_svc, features_columns_svc = load_ml_model(SVC_MODEL_PATH)
model_lgbm, scaler_lgbm, features_columns_lgbm = load_ml_model(LGBM_MODEL_PATH)
MODEL_MAPPING = {
'RandomForest': model_rf,
'SVC': model_svc,
'LGBM': model_lgbm,
}
SCALER_MAPPING = {
'RandomForest': scaler_rf,
'SVC': scaler_svc,
'LGBM': scaler_lgbm,
}
FEATURES_COLUMNS_MAPPING = {
'RandomForest': features_columns_rf,
'SVC': features_columns_svc,
'LGBM': features_columns_lgbm,
}
@st.cache(allow_output_mutation=True)
def load_audio(file_path):
y, sr = load_audio_file(file_path=file_path)
return y, sr
# Créer dataset à partir d'une liste de fichiers audio
@st.cache_resource(show_spinner=True)
def create_dataset_from_audio_files(audio_files: list):
# result dataset : columns = file_path, file_name
file_path_series = pd.Series(audio_files)
file_name_series = file_path_series.apply(lambda x: os.path.basename(x).split('.')[0])
dataset = pd.DataFrame({
'file_path': file_path_series,
'file_name': file_name_series,
})
return dataset
def make_archive(source, destination):
base = os.path.basename(destination)
name = base.split('.')[0]
format = base.split('.')[1]
archive_from = os.path.dirname(source)
archive_to = os.path.basename(source.strip(os.sep))
shutil.make_archive(name, format, archive_from, archive_to)
shutil.move('%s.%s' % (name, format), destination)
return destination
def main():
uploaded_file = st.file_uploader("Choose a ZIP file", type="zip")
if uploaded_file is not None:
with tempfile.TemporaryDirectory() as temp_dir:
# get the name of the uploaded file
uploaded_file_name = uploaded_file.name
# extract zip file
with zipfile.ZipFile(uploaded_file, 'r') as zip_ref:
zip_ref.extractall(temp_dir)
# create columns 20% - 80%
col1, col2 = st.columns([0.2, 0.8])
if temp_dir is not None:
# get all audios files in the dataset (wav, mp3, ...) use glob
audios = get_all_audios_files(temp_dir)
# display count of audios files
if len(audios) == 0:
st.error("No audio files found", icon='📂')
else:
col1.metric(label="📂 Audio files found.", value=len(audios))
model_name = col2.selectbox(
"Select the model 🤖",
MODEL_MAPPING.keys(),
)
# submit button
ok_button = col1.button("Predict", key="predict")
st.session_state.ok_submit = ok_button
if st.session_state.ok_submit and len(audios) > 0:
model = MODEL_MAPPING[model_name]
scaler = SCALER_MAPPING[model_name]
# get features from model
if model_name == 'RandomForest':
features_columns = model.feature_importances_
print(features_columns)
# create dataset from audio files
dataset = create_dataset_from_audio_files(audios)
# add features to dataset
with st.spinner("Extracting features..."):
dataset = extract_features_from_dataset(dataset, bool_streamlit=True)
with st.expander("Show dataset"):
# display dataset
# features selection : keep only features columns which are in the model
st.write("dataset", dataset.shape, dataset.columns.tolist())
st.dataframe(dataset.head())
X = dataset[FEATURES_COLUMNS_MAPPING[model_name]]
# scale dataset
X = scaler.transform(X)
st.write("X", X.shape)
st.write(X)
# make predictions and probabilities
with st.spinner("Making predictions..."):
predictions = model.predict(X)
probabilities = model.predict_proba(X) # get probabilities
# new code...
confidence_threshold = 0.75 # set your confidence threshold here
top_n = 2 # set the number of top predictions you want to show
# for each prediction, get the top 2 classes and their probabilities
top_predictions = []
for proba in probabilities:
# get top 2 predictions
top_indices = np.argsort(proba)[-top_n:]
top_probs = proba[top_indices]
top_classes = model.classes_[top_indices]
# if the highest probability is less than the threshold, show top 2 predictions
if top_probs[-1] < confidence_threshold:
top_prediction = f"{top_classes[-1]} ({top_probs[-1] * 100:.0f}%), {top_classes[-2]} ({top_probs[-2] * 100:.0f}%)"
else:
top_prediction = f"{top_classes[-1]} ({top_probs[-1] * 100:.0f}%)"
top_predictions.append(top_prediction)
# concatenate top_predictions to dataset
dataset['top_predictions'] = top_predictions
dataset['predictions'] = predictions
# display predictions in a grid
st.subheader("Predictions 🔮")
with st.expander("Show predictions"):
st.caption(f"Showing top {top_n} predictions with confidence > {confidence_threshold}")
st.caption(f"*file_name* ⇢ **top_prediction**")
n_cols = 2
cols = st.columns(n_cols)
for e, (idx_row, row) in enumerate(dataset.iterrows()):
audio_file = row['file_path']
audio_name = row['file_name']
top_prediction = row['top_predictions']
cols[e % n_cols].audio(audio_file)
cols[e % n_cols].markdown(f'"*{audio_name}*" ⇢ **{top_prediction}**')
st.subheader("Download 📥")
# buttons
# download dataset with predictions
with st.spinner("Downloading dataset with predictions..."):
name_download = f"predictions - {uploaded_file_name}.csv"
st.download_button(
label="Download dataset with predictions",
data=dataset.loc[:, ['file_name', 'predictions', 'top_predictions']].to_csv(index=False),
file_name=name_download,
mime="text/csv",
)
# Function to create classified directories
def create_classified_directories(temp_dir, dataset):
# verify if temp_dir exists
if not os.path.exists(temp_dir):
raise ValueError(f"temp_dir {temp_dir} does not exist")
# create directories
class_dirs = {class_name: os.path.join(temp_dir, class_name) for class_name in dataset['predictions'].unique()}
for class_name, class_dir in class_dirs.items():
if not os.path.exists(class_dir):
os.makedirs(class_dir)
st.write(class_dirs)
for idx, row in dataset.iterrows():
shutil.copy2(row['file_path'], os.path.join(class_dirs[row['predictions']], os.path.basename(row['file_path'])))
return temp_dir
# Functions to download files
def get_binary_file_downloader_html(bin_file, file_label='File'):
with open(bin_file, 'rb') as f:
data = f.read()
bin_str = base64.b64encode(data).decode()
href = f'<a href="data:application/octet-stream;base64,{bin_str}" download="{os.path.basename(bin_file)}">{file_label}</a>'
return href
if __name__ == "__main__":
main()