Spaces:
Running
Running
Create app.py
Browse files
app.py
ADDED
@@ -0,0 +1,34 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import gradio as gr
|
2 |
+
import time
|
3 |
+
from transformers import pipeline, AutoModelForSequenceClassification, AutoTokenizer
|
4 |
+
|
5 |
+
# ASR pipeline
|
6 |
+
asr_pipeline = pipeline("automatic-speech-recognition", model="facebook/wav2vec2-base-960h")
|
7 |
+
|
8 |
+
# Load classifier model and tokenizer
|
9 |
+
classifier_model = AutoModelForSequenceClassification.from_pretrained("Ngadou/bert-sms-spam-dectector")
|
10 |
+
classifier_tokenizer = AutoTokenizer.from_pretrained("Ngadou/bert-sms-spam-dectector")
|
11 |
+
|
12 |
+
def classify_audio(audio):
|
13 |
+
# Transcribe the audio to text
|
14 |
+
text = asr_pipeline(audio)["text"]
|
15 |
+
|
16 |
+
# Tokenize the text and feed it to the model
|
17 |
+
inputs = classifier_tokenizer.encode_plus(text, return_tensors="pt")
|
18 |
+
outputs = classifier_model(**inputs)
|
19 |
+
|
20 |
+
# Get the prediction (0 = ham, 1 = spam)
|
21 |
+
prediction = outputs.logits.argmax(dim=1).item()
|
22 |
+
|
23 |
+
# Return the transcription and the prediction as a dictionary
|
24 |
+
return text, "Scam" if prediction == 1 else "Safe Message"
|
25 |
+
|
26 |
+
gr.Interface(
|
27 |
+
fn=classify_audio,
|
28 |
+
inputs=gr.inputs.Audio(source="upload", type="filepath"),
|
29 |
+
outputs=[
|
30 |
+
gr.outputs.Textbox(label="Transcription"),
|
31 |
+
gr.outputs.Textbox(label="Classification"),
|
32 |
+
],
|
33 |
+
live=True
|
34 |
+
).launch(share=True)
|