Spaces:
Build error
Build error
| import gradio as gr | |
| import cv2 | |
| import librosa | |
| import numpy as np | |
| from keras.models import load_model | |
| import os | |
| import cv2 | |
| import json | |
| import pickle | |
| import librosa | |
| import shutil | |
| import numpy as np | |
| import pandas as pd | |
| from pathlib import Path | |
| from scipy.io import wavfile | |
| from moviepy.editor import VideoFileClip | |
| from keras.utils import np_utils | |
| from sklearn.preprocessing import LabelEncoder | |
| # ė°ģ“ķ° ģ ģ²ė¦¬ | |
| def preprocess_video(video_path): | |
| face_cascade = cv2.CascadeClassifier('/content/drive/Shareddrives/23 įį µį«įį ©į¼įį µįį ³į¼ įį ©įį ¦įÆį į µį¼_įį ©įÆįį µį«/haarcascade_frontalface_default.xml') | |
| cnn_data = [] | |
| rnn_data = [] | |
| cap = cv2.VideoCapture(video_path) | |
| count = 0 | |
| while len(cnn_data) < 2: | |
| ret, frame = cap.read() | |
| if ret: | |
| gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY) | |
| faces = face_cascade.detectMultiScale(gray, scaleFactor=1.1, minNeighbors=5) | |
| for (x, y, w, h) in faces: | |
| face_img = gray[y:y+h, x:x+w] | |
| resized_img = cv2.resize(face_img, (224, 224)) | |
| cnn_data.append(resized_img) | |
| count += 1 | |
| if count >= 15: | |
| break | |
| else: | |
| break | |
| if len(cnn_data) < 281: | |
| video_clip = VideoFileClip(video_path) | |
| audio_clip = video_clip.audio | |
| audio_clip.write_audiofile("audio.wav") | |
| y, sr = librosa.load("audio.wav", sr=44100) | |
| mfcc = librosa.feature.mfcc(y=y, sr=sr, n_mfcc=20) | |
| mfcc = mfcc[:, :400] | |
| rnn_data.append(mfcc) | |
| os.remove("audio.wav") | |
| cnn_data = np.array(cnn_data) | |
| rnn_data = np.array(rnn_data) | |
| return cnn_data, rnn_data | |
| # ė„ķģ“ķ¬ ģģ ģ ė¬“ ķė³ | |
| def detect_deepfake(video_path): | |
| cnn_data, rnn_data = preprocess_video(video_path) | |
| cnn_data_np = np.array(cnn_data) | |
| rnn_data_np= np.array(rnn_data) | |
| def augment_data(data, target_size): | |
| # ģ¦ź°ė ė°ģ“ķ° ė°°ģ“ ģ“źø°ķ | |
| augmented_data = np.empty((target_size,) + data.shape[1:]) | |
| # RNN ė°ģ“ķ°ė„¼ ė°ģ ķģ¬ ė³µģ¬ | |
| for i in range(target_size): | |
| augmented_data[i] = np.flip(data[i % data.shape[0]], axis=0) | |
| return augmented_data | |
| # RNN ė°ģ“ķ° ģ¦ź° | |
| augmented_rnn_data = augment_data(rnn_data_np, cnn_data_np.shape[0]) | |
| y_pred = multimodal_model.predict([cnn_data, augmented_rnn_data]) | |
| #print(y_pred) | |
| max_prob = np.max(y_pred) | |
| print(max_prob) | |
| if max_prob < 0.5: | |
| result = "Deepfake" | |
| else: | |
| result = "Real" | |
| return result | |
| iface = gr.Interface( | |
| fn=detect_deepfake, | |
| inputs="video", | |
| outputs="text", | |
| title="Video Deepfake Detection", | |
| description="Upload a video to check if it contains deepfake content.", | |
| allow_flagging=False, | |
| analytics_enabled=False | |
| ) | |
| iface.launch() |