app.py

import streamlit as st
import os
import shutil
import librosa
import numpy as np
from sklearn.ensemble import RandomForestClassifier
from sklearn.metrics import accuracy_score
import joblib
from finalpredicted import predict_deepfake

def extract_features(file_path):
    try:
        audio, sample_rate = librosa.load(file_path, res_type='kaiser_fast') 
        mfccs = np.mean(librosa.feature.mfcc(y=audio, sr=sample_rate, n_mfcc=40).T, axis=0)
        return mfccs
    except Exception as e:
        st.error(f"Error encountered while parsing file: {file_path}")
        return None

def classify_audio(example_file_path):
    import os

    current_dir = os.path.dirname(os.path.abspath(__file__))

    os.chdir(current_dir)

    loaded_model = joblib.load("svm_model.joblib")

    example_features = extract_features(example_file_path)
    if example_features is not None:
        prediction = loaded_model.predict([example_features])
        class_label = "Real" if prediction[0] == 1 else "Fake"
        return f"{class_label} Audio File"
    else:
        return "Error extracting features from the example file."
    
def check_video(uploaded_video_file, method):
    with st.spinner("Checking video..."):
        input_video_file_path = "uploaded_video.mp4"
        with open(input_video_file_path, "wb") as f:
            f.write(uploaded_video_file.getbuffer())
        fake_prob, real_prob, pred = predict_deepfake(input_video_file_path, method)
        
    return fake_prob, real_prob, pred

def main():
    st.title("VideoCop")
    st.write("video files deepfake detection.")
    
  
    st.header("Video Deepfake Detection")
    uploaded_video_file = st.file_uploader("Choose a video file", type=["mp4"], key="video_uploader")
    method_mapping = {"MTCNN": "plain_frames"}

    if uploaded_video_file is not None:
        selected_option = st.selectbox("Select method", list(method_mapping.keys()))
        st.video(uploaded_video_file)

        method = method_mapping[selected_option]

        if st.button("Check Video"):
            fake_prob, real_prob, pred = check_video(uploaded_video_file, method)

            if pred is None:
                st.error("Failed to detect DeepFakes in the video.")
            else:
                label = "real" if pred == 0 else "deepfaked"
                probability = real_prob if pred == 0 else fake_prob
                probability = round(probability * 100, 4)

                if pred == 0:
                    st.success(f"The video is {label}, with a probability of: {probability}%")
                    shutil.rmtree("./output")
                else:
                    st.error(f"The video is {label}, with a probability of: {probability}%")
                    shutil.rmtree("./output")

if __name__ == "__main__":
    main()