Spaces:
Running
Running
Commit
·
3b0b181
1
Parent(s):
ab02fe1
gender detection app
Browse files- app.py +120 -2
- requirements.txt +54 -0
- runtime.txt +1 -0
app.py
CHANGED
@@ -1,4 +1,122 @@
|
|
|
|
|
|
|
|
|
|
|
|
1 |
import streamlit as st
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2 |
|
3 |
-
|
4 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
"""
|
2 |
+
Streamlit application for real-time gender detection from audio input.
|
3 |
+
Uses wav2vec2 model to analyze voice and predict speaker gender.
|
4 |
+
"""
|
5 |
+
|
6 |
import streamlit as st
|
7 |
+
import pyaudio
|
8 |
+
import numpy as np
|
9 |
+
import torch
|
10 |
+
from transformers import AutoFeatureExtractor, AutoModelForAudioClassification
|
11 |
+
import logging
|
12 |
+
|
13 |
+
# Configure logging
|
14 |
+
logging.basicConfig(level=logging.INFO, format='%(asctime)s - %(levelname)s - %(message)s')
|
15 |
+
|
16 |
+
# Define audio stream parameters
|
17 |
+
FORMAT = pyaudio.paInt16 # 16-bit resolution
|
18 |
+
CHANNELS = 1 # Mono audio
|
19 |
+
RATE = 16000 # 16kHz sampling rate
|
20 |
+
CHUNK = 1024 # Number of frames per buffer
|
21 |
+
|
22 |
+
@st.cache_resource
|
23 |
+
def load_model():
|
24 |
+
"""
|
25 |
+
Load the wav2vec2 model and feature extractor for gender recognition.
|
26 |
+
|
27 |
+
Returns:
|
28 |
+
tuple: A tuple containing the feature extractor and the model.
|
29 |
+
"""
|
30 |
+
model_path = "alefiury/wav2vec2-large-xlsr-53-gender-recognition-librispeech"
|
31 |
+
# model_path = "./local-model"
|
32 |
+
feature_extractor = AutoFeatureExtractor.from_pretrained(model_path)
|
33 |
+
model = AutoModelForAudioClassification.from_pretrained(model_path)
|
34 |
+
model.eval()
|
35 |
+
logging.info("Model loaded successfully.")
|
36 |
+
return feature_extractor, model
|
37 |
+
|
38 |
+
st.title("Gender Detection")
|
39 |
+
|
40 |
+
# Initialize session state
|
41 |
+
if 'listening' not in st.session_state:
|
42 |
+
st.session_state['listening'] = False
|
43 |
+
if 'prediction' not in st.session_state:
|
44 |
+
st.session_state['prediction'] = ""
|
45 |
+
|
46 |
+
# Function to stop listening
|
47 |
+
def stop_listening():
|
48 |
+
"""Stop the audio stream and update session state to stop listening."""
|
49 |
+
if 'stream' in st.session_state:
|
50 |
+
logging.info("Stopping stream")
|
51 |
+
st.session_state['stream'].stop_stream()
|
52 |
+
st.session_state['stream'].close()
|
53 |
+
if 'audio' in st.session_state:
|
54 |
+
logging.info("Stopping audio")
|
55 |
+
st.session_state['audio'].terminate()
|
56 |
+
st.session_state['listening'] = False
|
57 |
+
st.session_state['prediction'] = "Stopped listening, click 'Start Listening' to start again."
|
58 |
+
st.rerun()
|
59 |
+
|
60 |
+
def start_listening():
|
61 |
+
"""Start the audio stream and continuously process audio for gender detection."""
|
62 |
+
placeholder = st.empty()
|
63 |
+
try:
|
64 |
+
placeholder.write("Loading model...")
|
65 |
+
feature_extractor, model = load_model()
|
66 |
+
audio = pyaudio.PyAudio()
|
67 |
+
stream = audio.open(format=FORMAT,
|
68 |
+
channels=CHANNELS,
|
69 |
+
rate=RATE,
|
70 |
+
input=True,
|
71 |
+
frames_per_buffer=CHUNK)
|
72 |
+
|
73 |
+
st.session_state['stream'] = stream
|
74 |
+
st.session_state['audio'] = audio
|
75 |
+
st.session_state['listening'] = True
|
76 |
+
st.session_state['prediction'] = "Listening........................"
|
77 |
+
placeholder.write("Listening for audio...")
|
78 |
+
|
79 |
+
while st.session_state['listening']:
|
80 |
+
audio_data = np.array([], dtype=np.float32)
|
81 |
+
|
82 |
+
for _ in range(int(RATE / CHUNK * 1.5)):
|
83 |
+
# Read audio chunk from the stream
|
84 |
+
data = stream.read(CHUNK, exception_on_overflow=False)
|
85 |
+
|
86 |
+
# Convert byte data to numpy array and normalize
|
87 |
+
chunk_data = np.frombuffer(data, dtype=np.int16).astype(np.float32) / 32768.0
|
88 |
+
audio_data = np.concatenate((audio_data, chunk_data))
|
89 |
+
|
90 |
+
# Check if there is significant sound
|
91 |
+
if np.max(np.abs(audio_data)) > 0.05: # Threshold for detecting sound
|
92 |
+
# Process the audio data
|
93 |
+
inputs = feature_extractor(audio_data, sampling_rate=RATE, return_tensors="pt", padding=True)
|
94 |
+
# Perform inference
|
95 |
+
with torch.no_grad():
|
96 |
+
logits = model(**inputs).logits
|
97 |
+
predicted_ids = torch.argmax(logits, dim=-1)
|
98 |
+
|
99 |
+
# Map predicted IDs to labels
|
100 |
+
predicted_label = model.config.id2label[predicted_ids.item()]
|
101 |
+
|
102 |
+
if predicted_label != st.session_state['prediction']:
|
103 |
+
st.session_state['prediction'] = predicted_label
|
104 |
+
# st.write(f"Detected Gender: {predicted_label}")
|
105 |
+
placeholder.write(f"Detected Gender: {predicted_label}")
|
106 |
+
else:
|
107 |
+
st.session_state['prediction'] = "---- No significant sound detected, skipping prediction. ----"
|
108 |
+
placeholder.empty()
|
109 |
+
placeholder.empty()
|
110 |
+
except Exception as e:
|
111 |
+
logging.error(f"An error occurred: {e}")
|
112 |
+
st.error(f"An error occurred: {e}")
|
113 |
+
stop_listening()
|
114 |
|
115 |
+
# Buttons to start and stop listening
|
116 |
+
col1, col2 = st.columns(2)
|
117 |
+
with col1:
|
118 |
+
if st.button("Start Listening"):
|
119 |
+
start_listening()
|
120 |
+
with col2:
|
121 |
+
if st.button("Stop Listening"):
|
122 |
+
stop_listening()
|
requirements.txt
ADDED
@@ -0,0 +1,54 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
altair==5.5.0
|
2 |
+
attrs==25.1.0
|
3 |
+
blinker==1.9.0
|
4 |
+
cachetools==5.5.1
|
5 |
+
certifi==2025.1.31
|
6 |
+
charset-normalizer==3.4.1
|
7 |
+
click==8.1.8
|
8 |
+
filelock==3.17.0
|
9 |
+
fsspec==2025.2.0
|
10 |
+
gitdb==4.0.12
|
11 |
+
GitPython==3.1.44
|
12 |
+
huggingface-hub==0.28.1
|
13 |
+
idna==3.10
|
14 |
+
Jinja2==3.1.5
|
15 |
+
jsonschema==4.23.0
|
16 |
+
jsonschema-specifications==2024.10.1
|
17 |
+
markdown-it-py==3.0.0
|
18 |
+
MarkupSafe==3.0.2
|
19 |
+
mdurl==0.1.2
|
20 |
+
mpmath==1.3.0
|
21 |
+
narwhals==1.26.0
|
22 |
+
networkx==3.4.2
|
23 |
+
numpy==2.2.3
|
24 |
+
packaging==24.2
|
25 |
+
pandas==2.2.3
|
26 |
+
pillow==11.1.0
|
27 |
+
protobuf==5.29.3
|
28 |
+
pyarrow==19.0.0
|
29 |
+
PyAudio==0.2.14
|
30 |
+
pydeck==0.9.1
|
31 |
+
Pygments==2.19.1
|
32 |
+
python-dateutil==2.9.0.post0
|
33 |
+
pytz==2025.1
|
34 |
+
PyYAML==6.0.2
|
35 |
+
referencing==0.36.2
|
36 |
+
regex==2024.11.6
|
37 |
+
requests==2.32.3
|
38 |
+
rich==13.9.4
|
39 |
+
rpds-py==0.22.3
|
40 |
+
safetensors==0.5.2
|
41 |
+
six==1.17.0
|
42 |
+
smmap==5.0.2
|
43 |
+
streamlit==1.42.0
|
44 |
+
sympy==1.13.1
|
45 |
+
tenacity==9.0.0
|
46 |
+
tokenizers==0.21.0
|
47 |
+
toml==0.10.2
|
48 |
+
torch==2.6.0
|
49 |
+
tornado==6.4.2
|
50 |
+
tqdm==4.67.1
|
51 |
+
transformers==4.48.3
|
52 |
+
typing_extensions==4.12.2
|
53 |
+
tzdata==2025.1
|
54 |
+
urllib3==2.3.0
|
runtime.txt
ADDED
@@ -0,0 +1 @@
|
|
|
|
|
1 |
+
python-3.10
|