| import streamlit as st |
| from transformers import AutoTokenizer, AutoModelForSequenceClassification |
| import whisper |
| import tempfile |
| import os |
| import torch |
| import sqlite3 |
| import bcrypt |
| from moviepy.editor import VideoFileClip |
| import subprocess |
| import imageio_ffmpeg |
|
|
| |
| conn = sqlite3.connect('users.db', check_same_thread=False) |
| cursor = conn.cursor() |
| cursor.execute('''CREATE TABLE IF NOT EXISTS users ( |
| username TEXT PRIMARY KEY, |
| password TEXT NOT NULL |
| )''') |
| conn.commit() |
|
|
| |
| def hash_password(password): |
| return bcrypt.hashpw(password.encode(), bcrypt.gensalt()) |
|
|
| def verify_password(password, hashed): |
| return bcrypt.checkpw(password.encode(), hashed) |
|
|
| def add_user(username, password): |
| hashed_pwd = hash_password(password) |
| try: |
| cursor.execute("INSERT INTO users (username, password) VALUES (?, ?)", (username, hashed_pwd)) |
| conn.commit() |
| return True |
| except: |
| return False |
|
|
| def authenticate_user(username, password): |
| cursor.execute("SELECT password FROM users WHERE username = ?", (username,)) |
| result = cursor.fetchone() |
| if result and verify_password(password, result[0]): |
| return True |
| return False |
|
|
| |
| def login_signup_page(): |
| st.set_page_config(page_title="Login | Hate Speech Classifier", layout="centered") |
| if 'page' not in st.session_state: |
| st.session_state.page = 'login' |
|
|
| st.markdown('<h1 style="text-align:center; color:#005f73;">🗣️ Smart Hate Speech Classifier</h1>', unsafe_allow_html=True) |
| st.markdown('<p style="text-align:center; color:#0a9396;">Please log in or create a new account to continue.</p>', unsafe_allow_html=True) |
|
|
| if st.session_state.page == 'login': |
| username = st.text_input("👤 Username") |
| password = st.text_input("🔑 Password", type="password") |
| if st.button("Login"): |
| if authenticate_user(username, password): |
| st.session_state.logged_in = True |
| st.session_state.username = username |
| st.session_state.page = "overview" |
| st.success("Login successful!") |
| st.rerun() |
| else: |
| st.error("Invalid credentials") |
| if st.button("New user? Create an account"): |
| st.session_state.page = 'register' |
|
|
| elif st.session_state.page == 'register': |
| new_user = st.text_input("👤 New Username") |
| new_pass = st.text_input("🔑 New Password", type="password") |
| if st.button("Create Account"): |
| if add_user(new_user, new_pass): |
| st.success("Account created! You can now log in.") |
| else: |
| st.error("Username already exists!") |
| if st.button("Already have an account? Login"): |
| st.session_state.page = 'login' |
|
|
| |
| if "logged_in" not in st.session_state: |
| st.session_state.logged_in = False |
| if not st.session_state.logged_in: |
| login_signup_page() |
| st.stop() |
|
|
| |
| ffmpeg_path = imageio_ffmpeg.get_ffmpeg_exe() |
| def custom_run(cmd, *args, **kwargs): |
| if cmd[0] == "ffmpeg": |
| cmd[0] = ffmpeg_path |
| return subprocess.run(cmd, *args, **kwargs) |
| import whisper.audio |
| whisper.audio.run = custom_run |
|
|
| |
| @st.cache_resource |
| def load_whisper_model(): |
| return whisper.load_model("tiny") |
|
|
| @st.cache_resource |
| def load_bert_model(): |
| model_name = "Hate-speech-CNERG/bert-base-uncased-hatexplain" |
| tokenizer = AutoTokenizer.from_pretrained(model_name) |
| model = AutoModelForSequenceClassification.from_pretrained(model_name) |
| model.eval() |
| return tokenizer, model |
|
|
| whisper_model = load_whisper_model() |
| tokenizer, classifier_model = load_bert_model() |
|
|
| |
| HATE_KEYWORDS = ["ugly", "stupid", "idiot", "hate", "kill", "trash","fuck you","bitch"] |
|
|
| def classify_text(text): |
| if any(word in text.lower() for word in HATE_KEYWORDS): |
| return "Hate Speech", 1.0 |
| inputs = tokenizer(text, return_tensors="pt", truncation=True, padding=True) |
| with torch.no_grad(): |
| outputs = classifier_model(**inputs) |
| probs = torch.nn.functional.softmax(outputs.logits, dim=1) |
| confidence, pred = torch.max(probs, dim=1) |
| label = "Hate Speech" if pred.item() == 1 else "Not Hate Speech" |
| return label, confidence.item() |
|
|
| def show_result(label, score): |
| st.markdown("### 🔍 Prediction Result:") |
| score_percent = f"{score * 100:.2f}%" |
| if label == "Hate Speech": |
| st.error(f"{label} ({score_percent} confident)") |
| else: |
| st.success(f"{label} ({score_percent} confident)") |
|
|
| |
| with st.sidebar: |
| st.title("📋 Navigation") |
| if st.button("📘 Project Overview"): |
| st.session_state.page = "overview" |
| if st.button("🎙️ Hate Speech Detector"): |
| st.session_state.page = "detector" |
|
|
| if 'page' not in st.session_state: |
| st.session_state.page = "overview" |
|
|
| |
| page = st.session_state.get("page", "overview") |
|
|
| if page == "overview": |
| st.title("🗣️ Smart Hate Speech Classifier Using BERT & Whisper") |
| st.markdown(""" |
| This AI-based project detects hate speech in: |
| - ✍️ Text Input |
| - 🔊 Audio Files |
| - 🎥 Video Uploads |
| |
| ### 🔧 Models Used: |
| - `OpenAI Whisper` for Speech-to-Text |
| - `HateXplain BERT` for Hate Speech Classification |
| """) |
|
|
| elif page == "detector": |
| st.title("🎙️ Hate Speech Detection") |
| input_mode = st.radio("Choose Input Type", ["Text", "Audio", "Video Upload"]) |
|
|
| if input_mode == "Text": |
| text_input = st.text_area("📝 Enter your message:") |
| if st.button("Classify Text"): |
| if text_input.strip(): |
| label, score = classify_text(text_input) |
| show_result(label, score) |
| else: |
| st.warning("⚠️ Please enter some text.") |
|
|
| elif input_mode == "Audio": |
| audio_file = st.file_uploader("📤 Upload Audio File:", type=["wav", "mp3"]) |
| if audio_file: |
| with tempfile.NamedTemporaryFile(delete=False, suffix=".wav") as temp_audio: |
| temp_audio.write(audio_file.read()) |
| audio_path = temp_audio.name |
| st.audio(audio_path) |
| result = whisper_model.transcribe(audio_path) |
| transcribed = result["text"] |
| st.success("📝 Transcribed Text:") |
| st.info(transcribed) |
| label, score = classify_text(transcribed) |
| show_result(label, score) |
| os.remove(audio_path) |
|
|
| elif input_mode == "Video Upload": |
| video_file = st.file_uploader("📤 Upload Video File:", type=["mp4", "mov", "avi"]) |
| if video_file: |
| with tempfile.NamedTemporaryFile(delete=False, suffix=".mp4") as temp_video: |
| temp_video.write(video_file.read()) |
| video_path = temp_video.name |
| st.video(video_path) |
| clip = VideoFileClip(video_path) |
| audio_path = tempfile.NamedTemporaryFile(delete=False, suffix=".wav").name |
| clip.audio.write_audiofile(audio_path) |
| result = whisper_model.transcribe(audio_path) |
| transcribed = result["text"] |
| st.success("📝 Transcribed Text:") |
| st.info(transcribed) |
| label, score = classify_text(transcribed) |
| show_result(label, score) |
| os.remove(video_path) |
| os.remove(audio_path) |
|
|
| st.markdown("---") |
| st.caption("Built with ❤️ using Streamlit, Whisper, and BERT.") |
|
|
