app_gradio.py

"""
Gradio UI application for Batik Classification using VGG16 model
"""
import gradio as gr
import torch
import torch.nn as nn
from torchvision import transforms, models
from PIL import Image
import json
import numpy as np
from typing import Tuple, List

# Global variables
model = None
class_names = []
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
transform = None


def load_model():
    """Load VGG16 model and configuration"""
    global model, class_names, transform

    try:
        # Load model configuration
        with open('model_config.json', 'r') as f:
            config = json.load(f)

        num_classes = config['num_classes']
        class_names = config['class_names']
        image_size = config.get('image_size', 224)

        # Initialize VGG16 model
        model = models.vgg16(weights=None)
        # Modify classifier to match saved model architecture
        # The saved model has classifier.3 as output layer (111 classes)
        model.classifier[3] = nn.Linear(4096, num_classes)
        # Remove layers after classifier.3
        model.classifier = nn.Sequential(*list(model.classifier.children())[:4])

        # Load trained weights
        checkpoint = torch.load('models/vgg16_batik_best.pth', map_location=device)

        # Check if checkpoint is a dict with 'model_state_dict' key or direct state_dict
        if isinstance(checkpoint, dict) and 'model_state_dict' in checkpoint:
            state_dict = checkpoint['model_state_dict']
        else:
            state_dict = checkpoint

        # Remove '_orig_mod.' prefix if present (from torch.compile)
        new_state_dict = {}
        for key, value in state_dict.items():
            if key.startswith('_orig_mod.'):
                new_key = key.replace('_orig_mod.', '')
                new_state_dict[new_key] = value
            else:
                new_state_dict[key] = value

        model.load_state_dict(new_state_dict)
        model = model.to(device)
        model.eval()        # Define image preprocessing
        transform = transforms.Compose([
            transforms.Resize((image_size, image_size)),
            transforms.ToTensor(),
            transforms.Normalize(mean=[0.485, 0.456, 0.406],
                               std=[0.229, 0.224, 0.225])
        ])

        print(f"✅ Model loaded successfully on {device}")
        print(f"📊 Number of classes: {num_classes}")

    except Exception as e:
        print(f"❌ Error loading model: {str(e)}")
        raise


def predict_single(image: Image.Image) -> Tuple[str, float]:
    """
    Predict single class for an image

    Args:
        image: PIL Image

    Returns:
        Tuple of (predicted_class, confidence)
    """
    try:
        # Preprocess image
        if image is None:
            return "Error: No image provided", 0.0

        # Convert to RGB if needed
        if image.mode != 'RGB':
            image = image.convert('RGB')

        # Transform and add batch dimension
        input_tensor = transform(image).unsqueeze(0).to(device)

        # Make prediction
        with torch.no_grad():
            outputs = model(input_tensor)
            probabilities = torch.nn.functional.softmax(outputs, dim=1)
            confidence, predicted = torch.max(probabilities, 1)

        predicted_class = class_names[predicted.item()]
        confidence_score = confidence.item() * 100  # Convert to percentage

        return predicted_class, confidence_score

    except Exception as e:
        return f"Error: {str(e)}", 0.0


def predict_top_k(image: Image.Image, k: int = 5) -> dict:
    """
    Predict top-k classes for an image

    Args:
        image: PIL Image
        k: Number of top predictions

    Returns:
        Dictionary of class names and their confidence scores
    """
    try:
        # Preprocess image
        if image is None:
            return {"Error": 1.0}

        # Convert to RGB if needed
        if image.mode != 'RGB':
            image = image.convert('RGB')

        # Transform and add batch dimension
        input_tensor = transform(image).unsqueeze(0).to(device)

        # Make prediction
        with torch.no_grad():
            outputs = model(input_tensor)
            probabilities = torch.nn.functional.softmax(outputs, dim=1)
            top_probs, top_indices = torch.topk(probabilities, min(k, len(class_names)), dim=1)

        # Format results as dictionary for Gradio
        results = {}
        for i in range(min(k, len(class_names))):
            class_name = class_names[top_indices[0][i].item()]
            confidence = top_probs[0][i].item()
            results[class_name] = float(confidence)

        return results

    except Exception as e:
        return {"Error": f"{str(e)}"}


def format_prediction(image: Image.Image) -> Tuple[str, dict]:
    """
    Format prediction output for Gradio interface

    Args:
        image: PIL Image

    Returns:
        Tuple of (formatted_text, top_k_dict)
    """
    try:
        if image is None:
            return "❌ Silakan upload gambar batik terlebih dahulu", {}

        # Get single prediction
        predicted_class, confidence = predict_single(image)

        # Get top-5 predictions
        top_k_results = predict_top_k(image, k=5)

        # Format main result
        result_text = f"""
## 🎯 Hasil Prediksi

**Motif Batik:** `{predicted_class}`
**Confidence:** `{confidence:.2f}%`

---

### 📊 Top 5 Prediksi:
"""

        for idx, (class_name, conf) in enumerate(list(top_k_results.items())[:5], 1):
            bar = "█" * int(conf * 20)  # Simple bar visualization
            result_text += f"\n{idx}. **{class_name}** - {conf*100:.2f}%  \n   {bar}"

        return result_text, top_k_results

    except Exception as e:
        return f"❌ Error: {str(e)}", {}


def get_model_info() -> str:
    """Get model information"""
    info = f"""
### 📋 Informasi Model

- **Arsitektur:** VGG16
- **Device:** {device}
- **Jumlah Kelas:** {len(class_names)}
- **Status:** ✅ Model siap digunakan

### 🎨 Kategori Batik:
Total {len(class_names)} motif batik dari berbagai daerah di Indonesia
"""
    return info


# Load model at startup
load_model()

# Create Gradio interface
with gr.Blocks(title="Batik Classification - VGG16", theme=gr.themes.Soft()) as demo:

    gr.Markdown("""
    # 🎨 Klasifikasi Motif Batik Indonesia
    ### Menggunakan Model VGG16 Deep Learning

    Upload gambar batik untuk mengetahui motif dan asalnya!
    """)

    with gr.Tabs():

        # Tab 1: Single Prediction
        with gr.Tab("🖼️ Prediksi Tunggal"):
            with gr.Row():
                with gr.Column():
                    input_image = gr.Image(
                        type="pil",
                        label="Upload Gambar Batik",
                        height=400
                    )
                    predict_btn = gr.Button("🔍 Prediksi", variant="primary", size="lg")

                    gr.Examples(
                        examples=[],  # Add example images if available
                        inputs=input_image,
                        label="Contoh Gambar (jika tersedia)"
                    )

                with gr.Column():
                    output_text = gr.Markdown(label="Hasil Prediksi")
                    output_label = gr.Label(
                        label="Top 5 Prediksi",
                        num_top_classes=5
                    )

            predict_btn.click(
                fn=format_prediction,
                inputs=input_image,
                outputs=[output_text, output_label]
            )

        # Tab 2: Batch Prediction
        with gr.Tab("📁 Prediksi Batch"):
            gr.Markdown("### Upload multiple gambar batik sekaligus")

            batch_input = gr.File(
                file_count="multiple",
                file_types=["image"],
                label="Upload Gambar (Multiple)"
            )
            batch_btn = gr.Button("🔍 Prediksi Semua", variant="primary")
            batch_output = gr.Dataframe(
                headers=["Filename", "Predicted Class", "Confidence (%)"],
                label="Hasil Prediksi Batch"
            )

            def predict_batch(files):
                """Predict multiple images"""
                if files is None or len(files) == 0:
                    return []

                results = []
                for file in files:
                    try:
                        image = Image.open(file.name)
                        pred_class, confidence = predict_single(image)
                        results.append([file.name.split('/')[-1], pred_class, f"{confidence:.2f}"])
                    except Exception as e:
                        results.append([file.name.split('/')[-1], "Error", str(e)])

                return results

            batch_btn.click(
                fn=predict_batch,
                inputs=batch_input,
                outputs=batch_output
            )

        # Tab 3: Model Info
        with gr.Tab("ℹ️ Info Model"):
            gr.Markdown(get_model_info())

            with gr.Accordion("📜 Daftar Semua Kelas Batik", open=False):
                class_list = "\n".join([f"{i+1}. {name}" for i, name in enumerate(class_names)])
                gr.Textbox(
                    value=class_list,
                    label=f"Total {len(class_names)} Kelas",
                    lines=20,
                    max_lines=30
                )

    gr.Markdown("""
    ---
    ### 📝 Cara Penggunaan:
    1. **Prediksi Tunggal:** Upload satu gambar batik dan klik tombol Prediksi
    2. **Prediksi Batch:** Upload beberapa gambar sekaligus untuk prediksi massal
    3. **Info Model:** Lihat informasi lengkap tentang model dan daftar kelas

    ### 💡 Tips:
    - Gunakan gambar dengan kualitas yang baik untuk hasil terbaik
    - Pastikan gambar menunjukkan motif batik dengan jelas
    - Model mendukung format JPG, PNG, dan format gambar umum lainnya
    """)


# Launch the app
if __name__ == "__main__":
    try:
        demo.launch(
            server_name="127.0.0.1",
            server_port=7860,
            share=False,  # Ubah ke True jika mau public link
            inbrowser=True,
            quiet=False
        )
    except Exception as e:
        print(f"Error launching Gradio: {e}")
        # Fallback: try simpler launch
        demo.launch()