app.py

import gradio as gr
import torch
from transformers import AutoImageProcessor, AutoModelForImageClassification
from PIL import Image
import numpy as np
import plotly.express as px
import plotly.graph_objects as go

# EuroSAT class names (10 land cover classes)
EUROSAT_CLASSES = [
    "AnnualCrop",
    "Forest", 
    "HerbaceousVegetation",
    "Highway",
    "Industrial",
    "Pasture",
    "PermanentCrop",
    "Residential",
    "River",
    "SeaLake"
]

# Class descriptions for better user understanding
CLASS_DESCRIPTIONS = {
    "AnnualCrop": "🌾 Agricultural land with annual crops",
    "Forest": "🌲 Dense forest areas with trees",
    "HerbaceousVegetation": "🌿 Areas with herbaceous vegetation",
    "Highway": "🛣️ Major roads and highway infrastructure",
    "Industrial": "🏭 Industrial areas and facilities",
    "Pasture": "🐄 Pasture land for livestock",
    "PermanentCrop": "🍇 Permanent crop areas (vineyards, orchards)",
    "Residential": "🏘️ Residential areas and neighborhoods",
    "River": "🏞️ Rivers and waterways",
    "SeaLake": "🏔️ Seas, lakes, and large water bodies"
}

class EuroSATClassifier:
    def __init__(self, model_name="Adilbai/EuroSAT-Swin"):
        self.model_name = model_name
        self.processor = None
        self.model = None
        self.load_model()
    
    def load_model(self):
        """Load the model and processor"""
        try:
            self.processor = AutoImageProcessor.from_pretrained(self.model_name)
            self.model = AutoModelForImageClassification.from_pretrained(self.model_name)
            self.model.eval()
            print(f"✅ Model {self.model_name} loaded successfully!")
        except Exception as e:
            print(f"❌ Error loading model: {e}")
            # Fallback to a generic model if the specific one fails
            self.processor = AutoImageProcessor.from_pretrained("microsoft/swin-tiny-patch4-window7-224")
            self.model = AutoModelForImageClassification.from_pretrained("microsoft/swin-tiny-patch4-window7-224")
    
    def predict(self, image):
        """Make prediction on the input image"""
        if image is None:
            return None, None, "Please upload an image first!"
        
        try:
            # Preprocess the image
            inputs = self.processor(images=image, return_tensors="pt")
            
            # Make prediction
            with torch.no_grad():
                outputs = self.model(**inputs)
                predictions = torch.nn.functional.softmax(outputs.logits, dim=-1)
            
            # Get top predictions
            probabilities = predictions[0].numpy()
            
            # Create results dictionary
            results = {}
            for i, class_name in enumerate(EUROSAT_CLASSES):
                if i < len(probabilities):
                    results[class_name] = float(probabilities[i])
                else:
                    results[class_name] = 0.0
            
            # Sort by confidence
            sorted_results = dict(sorted(results.items(), key=lambda x: x[1], reverse=True))
            
            # Get top prediction
            top_class = list(sorted_results.keys())[0]
            top_confidence = list(sorted_results.values())[0]
            
            # Create confidence plot
            confidence_plot = self.create_confidence_plot(sorted_results)
            
            # Format result text
            result_text = f"🎯 **Prediction: {top_class}**\n\n"
            result_text += f"📊 **Confidence: {top_confidence:.1%}**\n\n"
            result_text += f"📝 **Description: {CLASS_DESCRIPTIONS.get(top_class, 'Land cover classification')}**\n\n"
            result_text += "### Top 3 Predictions:\n"
            
            for i, (class_name, confidence) in enumerate(list(sorted_results.items())[:3]):
                result_text += f"{i+1}. **{class_name}**: {confidence:.1%}\n"
            
            return sorted_results, confidence_plot, result_text
            
        except Exception as e:
            error_msg = f"❌ Error during prediction: {str(e)}"
            return None, None, error_msg
    
    def create_confidence_plot(self, results):
        """Create a confidence plot using Plotly"""
        classes = list(results.keys())
        confidences = [results[cls] * 100 for cls in classes]
        
        # Create color scale - top prediction in green, others in blue gradient
        colors = ['#2E8B57' if i == 0 else f'rgba(70, 130, 180, {0.8 - i*0.1})' for i in range(len(classes))]
        
        fig = go.Figure(data=[
            go.Bar(
                x=confidences,
                y=classes,
                orientation='h',
                marker_color=colors,
                text=[f'{conf:.1f}%' for conf in confidences],
                textposition='inside',
                textfont=dict(color='white', size=12),
            )
        ])
        
        fig.update_layout(
            title={
                'text': "🎯 Classification Confidence Scores",
                'x': 0.5,
                'xanchor': 'center',
                'font': {'size': 16, 'color': '#2C3E50'}
            },
            xaxis_title="Confidence (%)",
            yaxis_title="Land Cover Classes",
            height=500,
            margin=dict(l=10, r=10, t=50, b=10),
            plot_bgcolor='rgba(248, 249, 250, 0.8)',
            paper_bgcolor='white',
            font=dict(family="Arial, sans-serif", size=12, color="#2C3E50"),
            xaxis=dict(
                gridcolor='rgba(128, 128, 128, 0.2)',
                showgrid=True,
                range=[0, 100]
            ),
            yaxis=dict(
                gridcolor='rgba(128, 128, 128, 0.2)',
                showgrid=True,
                autorange="reversed"  # Show highest confidence at top
            )
        )
        
        return fig

# Initialize the classifier
classifier = EuroSATClassifier()

def classify_image(image):
    """Main classification function for Gradio interface"""
    return classifier.predict(image)

def get_sample_images():
    """Return some sample image descriptions"""
    return """
    ### 🖼️ Try these types of satellite images:
    
    - **🌾 Agricultural fields** - Crop lands and farmland
    - **🌲 Forest areas** - Dense tree coverage
    - **🏘️ Residential zones** - Urban neighborhoods  
    - **🏭 Industrial sites** - Factories and industrial areas
    - **🛣️ Highway systems** - Major roads and intersections
    - **💧 Water bodies** - Rivers, lakes, and seas
    - **🌿 Natural vegetation** - Grasslands and natural areas
    
    Upload a satellite/aerial image to see the land cover classification!
    """

# Custom CSS for better styling
custom_css = """
.gradio-container {
    font-family: 'Segoe UI', Tahoma, Geneva, Verdana, sans-serif;
}

.main-header {
    text-align: center;
    background: linear-gradient(90deg, #667eea 0%, #764ba2 100%);
    color: white;
    padding: 2rem;
    border-radius: 10px;
    margin-bottom: 2rem;
}

.upload-area {
    border: 2px dashed #667eea;
    border-radius: 10px;
    padding: 2rem;
    text-align: center;
    background: rgba(102, 126, 234, 0.05);
}

.result-text {
    background: linear-gradient(135deg, #f5f7fa 0%, #c3cfe2 100%);
    padding: 1.5rem;
    border-radius: 10px;
    border-left: 4px solid #667eea;
}
"""

# Create the Gradio interface
with gr.Blocks(css=custom_css, title="🛰️ EuroSAT Land Cover Classifier") as demo:
    gr.HTML("""
        <div class="main-header">
            <h1>🛰️ EuroSAT Land Cover Classifier</h1>
            <p>Advanced satellite image classification using Swin Transformer</p>
            <p><strong>Model:</strong> Adilbai/EuroSAT-Swin | <strong>Dataset:</strong> EuroSAT (10 land cover classes)</p>
        </div>
    """)
    
    with gr.Row():
        with gr.Column(scale=1):
            gr.HTML("<h3>📤 Upload Satellite Image</h3>")
            image_input = gr.Image(
                label="Upload a satellite/aerial image",
                type="pil",
                height=400,
                elem_classes="upload-area"
            )
            
            classify_btn = gr.Button(
                "🔍 Classify Land Cover", 
                variant="primary",
                size="lg"
            )
            
            gr.HTML("<div style='margin-top: 2rem;'>")
            gr.Markdown(get_sample_images())
            gr.HTML("</div>")
        
        with gr.Column(scale=1):
            gr.HTML("<h3>📊 Classification Results</h3>")
            
            result_text = gr.Markdown(
                value="Upload an image and click 'Classify Land Cover' to see results!",
                elem_classes="result-text"
            )
            
            confidence_plot = gr.Plot(
                label="Confidence Scores",
            )
    
    # Hidden component to store raw results
    raw_results = gr.JSON(visible=False)
    
    # Event handlers
    classify_btn.click(
        fn=classify_image,
        inputs=[image_input],
        outputs=[raw_results, confidence_plot, result_text]
    )
    
    # Also trigger on image upload
    image_input.change(
        fn=classify_image,
        inputs=[image_input],
        outputs=[raw_results, confidence_plot, result_text]
    )
    
    # Footer
    gr.HTML("""
        <div style="text-align: center; margin-top: 3rem; padding: 2rem; background: #f8f9fa; border-radius: 10px;">
            <h4>🔬 About This Model</h4>
            <p>This classifier uses the <strong>Swin Transformer</strong> architecture trained on the <strong>EuroSAT dataset</strong>.</p>
            <p>The EuroSAT dataset contains <strong>27,000 satellite images</strong> from <strong>34 European countries</strong>, covering <strong>10 different land cover classes</strong>.</p>
            <p>Perfect for environmental monitoring, urban planning, and agricultural analysis! 🌍</p>
            <br>
            <p><strong>Model:</strong> <a href="https://huggingface.co/Adilbai/EuroSAT-Swin" target="_blank">Adilbai/EuroSAT-Swin</a></p>
        </div>
    """)

# Launch the app
if __name__ == "__main__":
    demo.launch(
        share=True,
        server_name="0.0.0.0",
        server_port=7860,
        show_error=True
    )