import numpy as np
from huggingface_hub import hf_hub_download
from tensorflow.keras.models import load_model
from PIL import Image
import gradio as gr

# ----------------------------
# تحميل الوزن من HuggingFace
# ----------------------------

REPO_ID = "MooseML/EfficientNet-Cancer-Detection"
MODEL_FILE = "efficientnet_cancer_model.h5"  # موجود فعليًا في Files

model_path = hf_hub_download(
    repo_id=REPO_ID,
    filename=MODEL_FILE
)

model = load_model(model_path)

# ----------------------------
# تحضير الصورة
# ----------------------------

def preprocess(img: Image.Image):
    img = img.resize((96, 96))
    img = np.array(img).astype("float32") / 255.0
    img = np.expand_dims(img, axis=0)
    return img

labels = ["No Tumor", "Tumor"]

# ----------------------------
# دالة التنبؤ
# ----------------------------

def predict(img):
    x = preprocess(img)
    probs = model.predict(x)[0]
    return {
        labels[0]: float(probs[0]),
        labels[1]: float(probs[1]),
    }

# ----------------------------
# واجهة Gradio
# ----------------------------

demo = gr.Interface(
    fn=predict,
    inputs=gr.Image(type="pil", label="Upload histopathology patch"),
    outputs=gr.Label(),
    title="EfficientNet Tumor Detection Demo",
    description="Model: MooseML/EfficientNet-Cancer-Detection",
)

demo.launch()