main.py

import torch
import torch.nn.functional as F
from facenet_pytorch import MTCNN, InceptionResnetV1
from PIL import Image
import cv2
from pytorch_grad_cam import GradCAM
from pytorch_grad_cam.utils.model_targets import ClassifierOutputTarget
from pytorch_grad_cam.utils.image import show_cam_on_image
from fastapi import FastAPI, File, UploadFile, Depends, HTTPException, Header
from fastapi.responses import JSONResponse
from io import BytesIO

app = FastAPI()

API_KEY = "c50dd5ady0uRL0rdnSaVyrArYaN161edb06af8"

def get_api_key(api_key: str = Header(...)):
    if api_key != API_KEY:
        raise HTTPException(status_code=403, detail="Could not validate credentials")

DEVICE = 'cuda:0' if torch.cuda.is_available() else 'cpu'

# Load MTCNN for face detection
mtcnn = MTCNN(select_largest=False, post_process=False, device=DEVICE).eval()

# Initialize the new model (InceptionResnetV1 with vggface2 and 3-class classification)
model = InceptionResnetV1(pretrained="vggface2", classify=True, num_classes=3, device=DEVICE)


DESTINATION_FILE_PATH = 'Model/3Class_32_epoch.pth'

# Load the model checkpoint
checkpoint = torch.load(DESTINATION_FILE_PATH, map_location=torch.device('cpu'))
model.to(DEVICE)
model.eval()

# Prediction function using the new model
def predict(input_image: Image.Image):
    """Predict the label of the input_image"""
    if input_image.mode == 'RGBA':
        input_image = input_image.convert('RGB')
    
    # Detect face in the image
    face = mtcnn(input_image)
    if face is None:
        raise Exception('No face detected')
    
    face = face.unsqueeze(0)  # Add batch dimension
    face = F.interpolate(face, size=(256, 256), mode='bilinear', align_corners=False)
    
    prev_face = face.squeeze(0).permute(1, 2, 0).cpu().detach().int().numpy()
    prev_face = prev_face.astype('uint8')

    face = face.to(DEVICE).to(torch.float32) / 255.0
    face_image_to_plot = face.squeeze(0).permute(1, 2, 0).cpu().detach().int().numpy()

    # Grad-CAM setup
    target_layers = [model.block8.branch1[-1]]
    cam = GradCAM(model=model, target_layers=target_layers)
    targets = [ClassifierOutputTarget(0), ClassifierOutputTarget(1), ClassifierOutputTarget(2)]

    grayscale_cam = cam(input_tensor=face, targets=targets, eigen_smooth=True)
    grayscale_cam = grayscale_cam[0, :]
    visualization = show_cam_on_image(face_image_to_plot, grayscale_cam, use_rgb=True)
    face_with_mask = cv2.addWeighted(prev_face, 1, visualization, 0.5, 0)

    # Inference
    with torch.no_grad():
        output = torch.softmax(model(face).squeeze(0), dim=0)
        class_indices = {0: 'real', 1: 'fake', 2: 'ai_generated'}
        prediction = class_indices[torch.argmax(output).item()]

        confidences = {
            'real': output[0].item(),
            'fake': output[1].item(),
            'ai_generated': output[2].item()
        }
    
    return confidences, prediction, face_with_mask

# FastAPI prediction endpoint
@app.post("/predict")
async def predict_api(file: UploadFile = File(...), api_key: str = Depends(get_api_key)):
    image = Image.open(BytesIO(await file.read()))
    try:
        confidences, prediction, face_with_mask = predict(image)
        _, buffer = cv2.imencode('.jpg', face_with_mask)
        face_with_mask_encoded = buffer.tobytes()
        return JSONResponse(content={
            "confidences": confidences,
            "prediction": prediction,
            "face_with_mask": face_with_mask_encoded.hex()
        })
    except Exception as e:
        return JSONResponse(content={"error": str(e)}, status_code=400)