import gradio as gr
import torch
from PIL import Image
from torchvision import transforms
import os
import numpy as np
import random
from resnet import resnet50

def seed_torch(seed=1029):
    random.seed(seed)
    os.environ['PYTHONHASHSEED'] = str(seed)
    np.random.seed(seed)
    torch.manual_seed(seed)
    torch.cuda.manual_seed(seed)
    torch.cuda.manual_seed_all(seed)
    torch.backends.cudnn.benchmark = False
    torch.backends.cudnn.deterministic = True
    torch.backends.cudnn.enabled = False

seed_torch(100)

def load_model(model_path):
    model = resnet50(num_classes=1)
    state_dict = torch.load(model_path, map_location='cpu')
    model.load_state_dict(state_dict, strict=True)
    if torch.cuda.is_available():
        model.cuda()
    model.eval()
    return model

def preprocess_image(image):
    transform = transforms.Compose([
        transforms.Resize((224, 224)),
        transforms.ToTensor(),
        transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])
    ])
    image = transform(image).unsqueeze(0)
    return image

def predict_image(model, image):
    if torch.cuda.is_available():
        image = image.cuda()
    with torch.no_grad():
        output = model(image)
        # Apply sigmoid to get probability between 0 and 1
        prediction = torch.sigmoid(output).item()
        
        # Clamp prediction between 0 and 1
        prediction = max(0, min(prediction, 1))
        
        # Convert to percentages
        real_prob = round(prediction * 1, 2)  # Rounded to 2 decimal places
        fake_prob = round(1 - real_prob, 2)  # Complementary probability
    
    return real_prob, fake_prob


# def predict_image(model, image):
#     if torch.cuda.is_available():
#         image = image.cuda()
#     with torch.no_grad():
#         output = model(image)
#         prediction = torch.sigmoid(output).item()
#         real_prob = gr.number(min(max(prediction * 100, 0), 100))  # Convert to integer
#         fake_prob = int(100 - real_prob)  # Ensure complementary probability
#     return real_prob, fake_prob

# Load the model once at the start
model_path = "model_epoch_last_3090.pth"  # Update with the correct path to your model
model = load_model(model_path)

def detect_deepfake(image):
    image = Image.fromarray(image).convert("RGB")
    preprocessed_image = preprocess_image(image)
    real_prob, fake_prob = predict_image(model, preprocessed_image)
    print("real_prob", real_prob)
    print("fake_prob", fake_prob)

    return {"Real Confidence": real_prob, "Fake Confidence": fake_prob}


iface = gr.Interface(
    fn=detect_deepfake,
    inputs=gr.Image(type="numpy", label="Upload Image"),
    outputs=gr.Label(num_top_classes=2, label="Confidence Scores"),
    title="Deepfake Detection",
    description="Upload an image to determine its confidence scores for being real or fake."
)

if __name__ == "__main__":
    iface.launch()