Upload 4 files

AI-Powered Bullying & Abuse Detection

app.py

@@ -0,0 +1,22 @@
+import streamlit as st
+from audio_processing import transcribe_audio
+from text_analysis import analyze_text
+
+st.title("🎙️ AI-Powered Bullying & Abuse Detection")
+
+uploaded_file = st.file_uploader("Upload an audio file", type=["wav", "mp3", "ogg"])
+
+if uploaded_file is not None:
+    st.audio(uploaded_file, format="audio/wav")
+
+    with st.spinner("Processing audio..."):
+        transcript = transcribe_audio(uploaded_file)
+
+    st.subheader("Transcribed Text")
+    st.write(transcript)
+
+    with st.spinner("Analyzing text..."):
+        result = analyze_text(transcript)
+
+    st.subheader("Analysis Result")
+    st.json(result)

audio_processing.py

@@ -0,0 +1,18 @@
+import whisper
+import tempfile
+from pydub import AudioSegment
+import os
+
+# Set FFmpeg path manually (if needed)
+AudioSegment.converter = r"C:\ffmpeg\bin\ffmpeg.exe"
+
+def transcribe_audio(audio_file):
+    model = whisper.load_model("base")
+
+    # Convert to WAV if necessary
+    with tempfile.NamedTemporaryFile(delete=False, suffix=".wav") as temp_wav:
+        audio = AudioSegment.from_file(audio_file)
+        audio.export(temp_wav.name, format="wav")
+
+        result = model.transcribe(temp_wav.name)
+        return result["text"]

requirements.txt

@@ -0,0 +1,8 @@
+streamlit 
+openai 
+transformers 
+torch 
+vosk 
+pydub 
+numpy
+openai-whisper

text_analysis.py

@@ -0,0 +1,22 @@
+from transformers import pipeline
+
+# Load the model (use a fine-tuned model for abuse detection)
+classifier = pipeline("text-classification", model="unitary/toxic-bert")
+
+def analyze_text(text):
+    results = classifier(text)
+    
+    # Convert to readable format
+    final_result = {
+        "bullying": any(res["label"] == "toxic" and res["score"] > 0.5 for res in results),
+        "threat": any(res["label"] == "threat" and res["score"] > 0.5 for res in results),
+        "scolding": any(res["label"] == "insult" and res["score"] > 0.5 for res in results),
+        "abuse": any(res["label"] in ["toxic", "severe_toxic"] and res["score"] > 0.6 for res in results),
+        "detailed_scores": results
+    }
+    # Make detected categories bold
+    for key in ["bullying", "threat", "scolding", "abuse"]:
+        if final_result[key]:
+            final_result[key] = f"**{key.upper()} DETECTED**"
+            
+    return final_result