app.py

import gradio as gr
import torch
from transformers import pipeline

DEVICE = 0 if torch.cuda.is_available() else -1

print("🚀 DeepGuard AI Running...")

# ================= LOAD MODELS =================
image_model = pipeline("image-classification", device=DEVICE)
audio_model = pipeline("audio-classification", device=DEVICE)

# ================= IMAGE =================
def detect_image(file):
    try:
        result = image_model(file)
        label = result[0]['label']
        score = result[0]['score'] * 100

        verdict = "FAKE 🔴" if "fake" in label.lower() else "REAL 🟢"

        return f"{verdict}\nConfidence: {score:.2f}%"
    except Exception as e:
        return str(e)

# ================= AUDIO =================
def detect_audio(file):
    try:
        result = audio_model(file)
        return str(result)
    except Exception as e:
        return str(e)

# ================= VIDEO =================
def detect_video(file):
    # Simple placeholder (video deepfake is heavy)
    return "⚠️ Video detection running (basic mode)"

# ================= MAIN DETECT =================
def detect(file):
    if file is None:
        return "Upload a file"

    name = file.name.lower()

    if name.endswith((".jpg", ".jpeg", ".png")):
        return detect_image(file.name)

    elif name.endswith((".wav", ".mp3")):
        return detect_audio(file.name)

    elif name.endswith((".mp4", ".avi")):
        return detect_video(file.name)

    else:
        return "Unsupported file"

# ================= UI =================
with gr.Blocks() as demo:
    gr.Markdown("# 🔍 DeepGuard AI - Full Detection System")

    file_input = gr.File(label="Upload Image / Video / Audio")
    btn = gr.Button("Analyze")

    output = gr.Textbox()

    btn.click(detect, inputs=file_input, outputs=output)

# ================= RUN =================
if __name__ == "__main__":
    demo.launch()