app.py

import gradio as gr
import json
from transformers import pipeline
import warnings
from transformers import ASTForAudioClassification, ASTFeatureExtractor

warnings.filterwarnings("ignore")
model = ASTForAudioClassification.from_pretrained("ATDI-Group/AST_finetuned_SIGIDwiki", subfolder='AST_finetuned_SIGIDwiki')
feature_extractor = ASTFeatureExtractor.from_pretrained("ATDI-Group/AST_finetuned_SIGIDwiki", subfolder='AST_finetuned_SIGIDwiki')
audio_classifier = pipeline("audio-classification", model=model, feature_extractor=feature_extractor)

def analyze_signal(audio_path):
    if audio_path is None:
        return {"Error": "No audio file provided. Please upload a file."}
        
    print(f"Analysing audio file: {audio_path}...")
    try:
        acoustic_results = audio_classifier(audio_path, top_k=3)
        detected_sounds = {item['label']: round(item['score'], 2) for item in acoustic_results}
        
        holistic_understanding = {
            "Detected Sounds": detected_sounds,
        }
        return holistic_understanding
    except Exception as e:
        return {"Error": f"Could not process the audio file. Details: {str(e)}"}

iface = gr.Interface(
    fn=analyze_signal,
    inputs=gr.Audio(type="filepath", label="Upload your Audio File (.wav, .mp3)"),
    outputs=gr.JSON(label="Analysis Result"),
    title="Audio Spectrogram Transformer finetuned on SIGID wiki for Radio Signal Classification ",
    description="""
    https://huggingface.co/ATDI-Group/AST_finetuned_SIGIDwiki
    """,
    examples=[["S5065-MSK.wav"]] 
)

# This will create the web UI
iface.launch(debug=True)