from pathlib import Path

import gradio as gr
import numpy as np
from PIL import Image
from tensorflow import keras


PROJECT_ROOT = Path(__file__).resolve().parent
MODEL_PATH = PROJECT_ROOT / "face_mask_model.keras"
IMAGE_SIZE = (224, 224)
CLASS_NAMES = ["with_mask", "without_mask"]


if not MODEL_PATH.exists():
    raise FileNotFoundError(
        f"Model tidak ditemukan di {MODEL_PATH}. Pastikan file .keras tersedia."
    )


model = keras.models.load_model(MODEL_PATH)


def build_examples() -> list[list[str]]:
    example_paths = []
    candidate_files = [
        PROJECT_ROOT / "data" / "with_mask" / "with_mask_1.jpg",
        PROJECT_ROOT / "data" / "without_mask" / "without_mask_1.jpg",
        PROJECT_ROOT / "data" / "with_mask" / "with_mask_10.jpg",
        PROJECT_ROOT / "data" / "without_mask" / "without_mask_10.jpg",
    ]

    for path in candidate_files:
        if path.exists():
            example_paths.append([str(path)])

    return example_paths


def predict_mask(image: Image.Image):
    if image is None:
        raise gr.Error("Silakan unggah gambar wajah terlebih dahulu.")

    image = image.convert("RGB").resize(IMAGE_SIZE)
    image_array = np.asarray(image, dtype=np.float32)
    batch = np.expand_dims(image_array, axis=0)

    predictions = model.predict(batch, verbose=0)[0]
    scores = {label: float(score) for label, score in zip(CLASS_NAMES, predictions)}

    predicted_index = int(np.argmax(predictions))
    predicted_label = CLASS_NAMES[predicted_index]
    confidence = float(predictions[predicted_index]) * 100

    summary = (
        f"Prediksi: {predicted_label.replace('_', ' ').title()} | "
        f"Confidence: {confidence:.2f}%"
    )

    return scores, summary


custom_css = """
body {
    background:
        radial-gradient(circle at top left, #fff5d6 0%, transparent 32%),
        radial-gradient(circle at top right, #d8f5ff 0%, transparent 30%),
        linear-gradient(135deg, #f7f2e8 0%, #eef7ff 48%, #f5efe4 100%);
}
.gradio-container {
    max-width: 1100px !important;
}
.hero {
    padding: 20px 24px;
    border-radius: 24px;
    background: linear-gradient(135deg, rgba(17, 24, 39, 0.94), rgba(31, 41, 55, 0.88));
    color: #f9fafb;
    box-shadow: 0 16px 40px rgba(15, 23, 42, 0.15);
    margin-bottom: 16px;
}
.hero h1 {
    margin: 0 0 8px 0;
    font-size: 2.2rem;
}
.hero p {
    margin: 0;
    color: #e5e7eb;
    line-height: 1.6;
}
"""


with gr.Blocks(
    theme=gr.themes.Soft(
        primary_hue="cyan",
        secondary_hue="amber",
        neutral_hue="slate",
    ),
    css=custom_css,
    title="Face Mask Detector",
) as demo:
    gr.HTML(
        """
        <div class="hero">
            <h1>Face Mask Detector</h1>
            <p>
                Unggah gambar wajah untuk mendeteksi apakah orang tersebut memakai masker
                atau tidak. Model diambil langsung dari file <code>face_mask_model.keras</code>.
            </p>
        </div>
        """
    )

    with gr.Row():
        with gr.Column(scale=5):
            image_input = gr.Image(
                type="pil",
                label="Input Wajah",
                sources=["upload", "webcam"],
            )
            predict_button = gr.Button("Prediksi Sekarang", variant="primary")

        with gr.Column(scale=4):
            label_output = gr.Label(
                label="Probabilitas Kelas",
                num_top_classes=2,
            )
            summary_output = gr.Textbox(
                label="Ringkasan Hasil",
                interactive=False,
            )

    examples = build_examples()
    if examples:
        gr.Examples(
            examples=examples,
            inputs=image_input,
            label="Contoh Gambar",
        )

    predict_button.click(
        fn=predict_mask,
        inputs=image_input,
        outputs=[label_output, summary_output],
    )
    image_input.change(
        fn=predict_mask,
        inputs=image_input,
        outputs=[label_output, summary_output],
    )


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