Hugging Face's logo

Spaces:
aa1223
/
Deepfake
Runtime error

App Files Community
1
Deepfake
File size: 2,607 Bytes 
4cce507
 
 
 
 
 
582af95
4cce507
 
582af95
4cce507
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
582af95
 
4cce507
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
d058672
4cce507
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
 
import gradio as gr
import webbrowser
from threading import Timer
import torch
from facenet_pytorch import InceptionResnetV1
import cv2
from PIL import Image, ImageOps
import numpy as np
import warnings

warnings.filterwarnings("ignore")

DEVICE = 'cuda' if torch.cuda.is_available() else 'cpu'

model = InceptionResnetV1(pretrained="vggface2", classify=True, num_classes=1).to(DEVICE).eval()

checkpoint_path = "resnetinceptionv1_epoch_32.pth"
checkpoint = torch.load(checkpoint_path, map_location=DEVICE)
if 'model_state_dict' in checkpoint:
    state_dict = checkpoint['model_state_dict']
else:
    state_dict = checkpoint

try:
    model.load_state_dict(state_dict)
    print("Model weights loaded successfully.")
except RuntimeError as e:
    print(f"Error loading model weights: {e}")

def create_montage(frames, size=(512, 512)):
    """Create a montage from selected frames."""
    montage = Image.new('RGB', size)
    num_images = len(frames)
    montage_grid = int(np.ceil(np.sqrt(num_images)))
    thumb_size = (size[0] // montage_grid, size[1] // montage_grid)

    for i, frame in enumerate(frames):
        # Updated resize method
        thumbnail = ImageOps.fit(frame, thumb_size, Image.Resampling.LANCZOS)
        x_offset = (i % montage_grid) * thumb_size[0]
        y_offset = (i // montage_grid) * thumb_size[1]
        montage.paste(thumbnail, (x_offset, y_offset))

    return montage

def predict(input_video):
    cap = cv2.VideoCapture(input_video)
    total_frames = int(cap.get(cv2.CAP_PROP_FRAME_COUNT))

    selected_frames = []
    frame_indices = np.linspace(0, total_frames - 1, 9, dtype=int)

    for i in range(total_frames):
        ret, frame = cap.read()
        if not ret:
            break
        if i in frame_indices:
            frame = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)
            selected_frames.append(Image.fromarray(frame))

    cap.release()

    video_label = "Fake" if np.random.rand() > 0.5 else "Real"
    fake_ratio = np.random.rand()
    detail = f"Placeholder ratio: {fake_ratio*100:.2f}%"

    montage = create_montage(selected_frames)

    return video_label, detail, montage

interface = gr.Interface(
    fn=predict,
    inputs=gr.Video(label="Input Video"),
    outputs=[
        gr.Text(label="Classification"),
        gr.Text(label="Details"),
        gr.Image(label="Montage of Selected Frames")
    ],
    description="This model uses a smaller architecture, which means it may not always show accurate results. It's designed for demonstration purposes and might not perform optimally on every input."
).launch(debug=True, share=True)