app.py

import streamlit as st
import tensorflow as tf
import numpy as np
import pandas as pd
import plotly.express as px
import plotly.graph_objects as go
from plotly.subplots import make_subplots
from PIL import Image
import requests
import io
from datetime import datetime
import time
import logging

# Set up logging to help debug issues
logging.basicConfig(level=logging.INFO)
logger = logging.getLogger(__name__)

# Configure page - move this to the very top and add error handling
try:
    st.set_page_config(
        page_title="Satellite Classification Dashboard",
        page_icon="🛰️",
        layout="wide",
        initial_sidebar_state="expanded"
    )
except Exception as e:
    logger.error(f"Error setting page config: {e}")
    # Continue without custom config if it fails

# Custom CSS for better styling
st.markdown("""
<style>
    .main-header {
        font-size: 3rem;
        font-weight: bold;
        text-align: center;
        color: #1f77b4;
        margin-bottom: 2rem;
    }
    .model-card {
        background: linear-gradient(135deg, #667eea 0%, #764ba2 100%);
        padding: 20px;
        border-radius: 10px;
        margin: 10px 0;
        color: white;
    }
    .metric-card {
        background: #f8f9fa;
        padding: 15px;
        border-radius: 8px;
        border-left: 4px solid #1f77b4;
        margin: 5px 0;
    }
    .prediction-box {
        background: linear-gradient(135deg, #ff7e5f 0%, #feb47b 100%);
        padding: 20px;
        border-radius: 10px;
        text-align: center;
        color: white;
        font-size: 1.2rem;
    }
    .stAlert > div {
        padding: 10px;
        border-radius: 5px;
    }
</style>
""", unsafe_allow_html=True)

# Class mappings
CLASS_NAMES = {
    0: 'AcrimSat', 1: 'Aquarius', 2: 'Aura', 3: 'Calipso', 4: 'Cloudsat',
    5: 'CubeSat', 6: 'Debris', 7: 'Jason', 8: 'Sentinel-6', 9: 'TRMM', 10: 'Terra'
}

# Model configurations
MODEL_CONFIGS = {
    "Custom CNN": {
        "url": "https://huggingface.co/Bhavi23/Custom_CNN/resolve/main/best_multimodal_model.keras",
        "description": "Custom CNN architecture designed for satellite classification",
        "input_shape": (224, 224, 3),
        "strengths": ["Good generalization", "Balanced performance", "Stable training"],
        "best_for": ["General purpose", "Balanced datasets", "When interpretability matters"]
    },
    "MobileNetV2": {
        "url": "https://huggingface.co/Bhavi23/MobilenetV2/resolve/main/multi_input_model_v1.keras",
        "description": "Lightweight model optimized for mobile deployment",
        "input_shape": (224, 224, 3),
        "strengths": ["Fast inference", "Small model size", "Energy efficient"],
        "best_for": ["Real-time applications", "Mobile devices", "Resource constraints"]
    },
    "EfficientNetB0": {
        "url": "https://huggingface.co/Bhavi23/EfficientNet_B0/resolve/main/efficientnet_model.keras",
        "description": "Balanced efficiency and accuracy with compound scaling",
        "input_shape": (224, 224, 3),
        "strengths": ["High accuracy", "Parameter efficient", "Good transfer learning"],
        "best_for": ["High accuracy needs", "Limited data", "Transfer learning scenarios"]
    },
    "DenseNet121": {
        "url": "https://huggingface.co/Bhavi23/DenseNet/resolve/main/densenet_model.keras",
        "description": "Dense connections for feature reuse and gradient flow",
        "input_shape": (224, 224, 3),
        "strengths": ["Feature reuse", "Good gradient flow", "Parameter efficiency"],
        "best_for": ["Complex patterns", "Feature-rich images", "When accuracy is priority"]
    }
}

# Performance metrics (based on the results shown in your document)
MODEL_METRICS = {
    "Custom CNN": {
        "accuracy": 95.2,
        "precision": 94.8,
        "recall": 95.1,
        "f1_score": 94.9,
        "inference_time": 45,  # ms
        "model_size": 25.3,    # MB
        "training_time": 120   # minutes
    },
    "MobileNetV2": {
        "accuracy": 92.8,
        "precision": 92.1,
        "recall": 92.5,
        "f1_score": 92.3,
        "inference_time": 18,  # ms
        "model_size": 8.7,     # MB
        "training_time": 95    # minutes
    },
    "EfficientNetB0": {
        "accuracy": 96.4,
        "precision": 96.1,
        "recall": 96.2,
        "f1_score": 96.1,
        "inference_time": 35,  # ms
        "model_size": 20.1,    # MB
        "training_time": 140   # minutes
    },
    "DenseNet121": {
        "accuracy": 94.7,
        "precision": 94.2,
        "recall": 94.5,
        "f1_score": 94.3,
        "inference_time": 52,  # ms
        "model_size": 32.8,    # MB
        "training_time": 160   # minutes
    }
}

@st.cache_resource
def load_model(model_name):
    """Load model from HuggingFace with caching and better error handling"""
    try:
        logger.info(f"Loading model: {model_name}")
        url = MODEL_CONFIGS[model_name]["url"]
        
        # Add timeout and better error handling
        response = requests.get(url, timeout=60, stream=True)
        response.raise_for_status()
        
        # Check if response is actually a Keras model
        if len(response.content) < 1000:  # Too small to be a model
            st.error(f"Model {model_name} download failed - file too small")
            return None
            
        model_bytes = io.BytesIO(response.content)
        
        # Try to load the model with error handling
        try:
            model = tf.keras.models.load_model(model_bytes)
            logger.info(f"Successfully loaded model: {model_name}")
            return model
        except Exception as load_error:
            st.error(f"Error loading Keras model {model_name}: {str(load_error)}")
            return None
            
    except requests.exceptions.Timeout:
        st.error(f"Timeout loading {model_name}. Please try again.")
        return None
    except requests.exceptions.RequestException as e:
        st.error(f"Network error loading {model_name}: {str(e)}")
        return None
    except Exception as e:
        st.error(f"Unexpected error loading {model_name}: {str(e)}")
        logger.error(f"Error loading {model_name}: {str(e)}")
        return None

def preprocess_image(image, target_size=(224, 224)):
    """Preprocess image for model prediction with error handling"""
    try:
        if image.mode != 'RGB':
            image = image.convert('RGB')
        image = image.resize(target_size)
        image_array = np.array(image) / 255.0
        return np.expand_dims(image_array, axis=0)
    except Exception as e:
        st.error(f"Error preprocessing image: {str(e)}")
        return None

def predict_with_model(model, image, model_name):
    """Make prediction with a specific model with better error handling"""
    if model is None:
        return None
        
    try:
        start_time = time.time()
        predictions = model.predict(image, verbose=0)
        inference_time = (time.time() - start_time) * 1000  # Convert to ms
        
        # Validate predictions
        if predictions is None or len(predictions) == 0:
            st.error(f"No predictions returned from {model_name}")
            return None
            
        predicted_class = np.argmax(predictions[0])
        confidence = np.max(predictions[0]) * 100
        
        # Validate class prediction
        if predicted_class not in CLASS_NAMES:
            st.error(f"Invalid class prediction from {model_name}: {predicted_class}")
            return None
        
        return {
            'class': predicted_class,
            'class_name': CLASS_NAMES[predicted_class],
            'confidence': confidence,
            'inference_time': inference_time,
            'probabilities': predictions[0]
        }
    except Exception as e:
        st.error(f"Prediction error with {model_name}: {str(e)}")
        logger.error(f"Prediction error with {model_name}: {str(e)}")
        return None

def recommend_best_model(image_predictions):
    """Recommend the best model based on predictions and confidence"""
    if not image_predictions:
        return "EfficientNetB0"  # Default recommendation
    
    # Calculate recommendation score based on confidence and model performance
    recommendations = {}
    for model_name, pred in image_predictions.items():
        if pred:
            # Combine confidence with model's overall accuracy
            base_score = MODEL_METRICS[model_name]["accuracy"]
            confidence_bonus = pred['confidence'] * 0.1
            speed_bonus = max(0, 100 - MODEL_METRICS[model_name]["inference_time"]) * 0.05
            
            recommendations[model_name] = base_score + confidence_bonus + speed_bonus
    
    if recommendations:
        best_model = max(recommendations, key=recommendations.get)
        return best_model
    return "EfficientNetB0"

def create_metrics_comparison():
    """Create comprehensive metrics comparison dashboard with error handling"""
    try:
        # Create subplots
        fig = make_subplots(
            rows=2, cols=2,
            subplot_titles=('Accuracy Comparison', 'Model Size vs Inference Time', 
                           'Performance Metrics Radar', 'Training Efficiency'),
            specs=[[{"type": "bar"}, {"type": "scatter"}],
                   [{"type": "scatterpolar"}, {"type": "bar"}]]
        )
        
        models = list(MODEL_METRICS.keys())
        
        # 1. Accuracy Comparison Bar Chart
        accuracies = [MODEL_METRICS[model]["accuracy"] for model in models]
        fig.add_trace(
            go.Bar(x=models, y=accuracies, name="Accuracy", 
                   marker_color=['#1f77b4', '#ff7f0e', '#2ca02c', '#d62728']),
            row=1, col=1
        )
        
        # 2. Model Size vs Inference Time Scatter
        sizes = [MODEL_METRICS[model]["model_size"] for model in models]
        times = [MODEL_METRICS[model]["inference_time"] for model in models]
        fig.add_trace(
            go.Scatter(x=sizes, y=times, mode='markers+text', 
                       text=models, textposition="top center",
                       marker=dict(size=15, color=['#1f77b4', '#ff7f0e', '#2ca02c', '#d62728']),
                       name="Size vs Speed"),
            row=1, col=2
        )
        
        # 3. Radar Chart for Performance Metrics
        metrics = ['accuracy', 'precision', 'recall', 'f1_score']
        colors = ['#1f77b4', '#ff7f0e', '#2ca02c', '#d62728']
        for i, model in enumerate(models):
            values = [MODEL_METRICS[model][metric] for metric in metrics]
            fig.add_trace(
                go.Scatterpolar(r=values, theta=metrics, fill='toself', 
                               name=model, opacity=0.7, line_color=colors[i]),
                row=2, col=1
            )
        
        # 4. Training Time Comparison
        training_times = [MODEL_METRICS[model]["training_time"] for model in models]
        fig.add_trace(
            go.Bar(x=models, y=training_times, name="Training Time",
                   marker_color=['#9467bd', '#8c564b', '#e377c2', '#7f7f7f']),
            row=2, col=2
        )
        
        # Update layout
        fig.update_layout(height=800, showlegend=True, 
                         title_text="Comprehensive Model Comparison Dashboard")
        fig.update_xaxes(title_text="Models", row=1, col=1)
        fig.update_yaxes(title_text="Accuracy (%)", row=1, col=1)
        fig.update_xaxes(title_text="Model Size (MB)", row=1, col=2)
        fig.update_yaxes(title_text="Inference Time (ms)", row=1, col=2)
        fig.update_xaxes(title_text="Models", row=2, col=2)
        fig.update_yaxes(title_text="Training Time (minutes)", row=2, col=2)
        
        return fig
    except Exception as e:
        st.error(f"Error creating metrics comparison chart: {str(e)}")
        return None

def create_class_distribution_chart():
    """Create class distribution visualization with error handling"""
    try:
        classes = list(CLASS_NAMES.values())
        samples = [7500 if cls != 'Debris' else 15000 for cls in classes]
        percentages = [8.33 if cls != 'Debris' else 16.67 for cls in classes]
        
        fig = go.Figure()
        fig.add_trace(go.Bar(
            x=classes,
            y=samples,
            text=[f'{s} ({p:.1f}%)' for s, p in zip(samples, percentages)],
            textposition='auto',
            marker_color=['#ff6b6b' if cls == 'Debris' else '#4ecdc4' for cls in classes]
        ))
        
        fig.update_layout(
            title="Class Distribution in Training Dataset",
            xaxis_title="Satellite Classes",
            yaxis_title="Number of Samples",
            height=400,
            xaxis_tickangle=-45
        )
        
        return fig
    except Exception as e:
        st.error(f"Error creating class distribution chart: {str(e)}")
        return None

def create_confusion_matrix_heatmap():
    """Create a sample confusion matrix heatmap for demonstration"""
    try:
        # Sample confusion matrix data (you would replace this with actual data)
        classes = list(CLASS_NAMES.values())
        np.random.seed(42)  # For reproducible demo data
        
        # Create a realistic-looking confusion matrix
        confusion_matrix = np.random.randint(0, 100, size=(len(classes), len(classes)))
        # Make diagonal elements higher (correct predictions)
        np.fill_diagonal(confusion_matrix, np.random.randint(400, 500, size=len(classes)))
        
        fig = go.Figure(data=go.Heatmap(
            z=confusion_matrix,
            x=classes,
            y=classes,
            colorscale='Blues',
            showscale=True
        ))
        
        fig.update_layout(
            title="Sample Confusion Matrix (Demo Data)",
            xaxis_title="Predicted Class",
            yaxis_title="True Class",
            height=600
        )
        
        return fig
    except Exception as e:
        st.error(f"Error creating confusion matrix: {str(e)}")
        return None

# Main App
def main():
    try:
        # Header
        st.markdown('<h1 class="main-header">🛰️ Satellite Classification Dashboard</h1>', 
                    unsafe_allow_html=True)
        
        # Sidebar
        st.sidebar.title("Navigation")
        page = st.sidebar.selectbox("Choose a page", 
                                   ["🏠 Home", "📊 Model Comparison", "🔍 Image Classification", 
                                    "📈 Performance Analytics", "ℹ️ About Models"])
        
        # Add sidebar information
        st.sidebar.markdown("---")
        st.sidebar.markdown("### System Info")
        st.sidebar.info(f"Total Classes: {len(CLASS_NAMES)}")
        st.sidebar.info(f"Available Models: {len(MODEL_CONFIGS)}")
        st.sidebar.info("Built with Streamlit & TensorFlow")
        
        if page == "🏠 Home":
            st.markdown("## Welcome to the Satellite Classification System")
            
            col1, col2 = st.columns(2)
            
            with col1:
                st.markdown("### 🎯 System Overview")
                st.write("""
                This dashboard provides comprehensive satellite classification using 4 different
                deep learning models. Upload satellite images to classify them into 11 different
                categories including various satellites and space debris.
                """)
                
                st.markdown("### 🛰️ Supported Classes")
                for i, (class_id, class_name) in enumerate(CLASS_NAMES.items()):
                    if i < 6:  # First column
                        st.write(f"• **{class_name}**")
            
            with col2:
                st.markdown("### 🤖 Available Models")
                st.write("""
                - **Custom CNN**: Tailored architecture for satellite imagery
                - **MobileNetV2**: Lightweight and fast inference
                - **EfficientNetB0**: Best accuracy-efficiency balance
                - **DenseNet121**: Complex pattern recognition
                """)
                
                st.markdown("### 📊 Remaining Classes")
                for i, (class_id, class_name) in enumerate(CLASS_NAMES.items()):
                    if i >= 6:  # Second column
                        st.write(f"• **{class_name}**")
            
            # Class distribution chart
            chart = create_class_distribution_chart()
            if chart:
                st.plotly_chart(chart, use_container_width=True)
                
            # Quick start guide
            st.markdown("### 🚀 Quick Start Guide")
            st.markdown("""
            1. Navigate to **🔍 Image Classification** to upload and classify satellite images
            2. Check **📊 Model Comparison** to compare different model performances
            3. Explore **📈 Performance Analytics** for detailed metrics
            4. Read **ℹ️ About Models** to understand each model's capabilities
            """)
        
        elif page == "📊 Model Comparison":
            st.markdown("## 📊 Model Performance Comparison")
            
            # Metrics table
            st.markdown("### Performance Metrics Summary")
            df_metrics = pd.DataFrame(MODEL_METRICS).T
            st.dataframe(df_metrics.style.highlight_max(axis=0), use_container_width=True)
            
            # Comprehensive comparison chart
            chart = create_metrics_comparison()
            if chart:
                st.plotly_chart(chart, use_container_width=True)
            
            # Model recommendations
            st.markdown("### 🎯 Model Selection Guide")
            
            col1, col2 = st.columns(2)
            
            with col1:
                st.markdown("#### 🏆 Best for Accuracy")
                st.success("**EfficientNetB0** - 96.4% accuracy")
                
                st.markdown("#### ⚡ Best for Speed")
                st.info("**MobileNetV2** - 18ms inference time")
            
            with col2:
                st.markdown("#### 💾 Most Lightweight")
                st.info("**MobileNetV2** - 8.7MB model size")
                
                st.markdown("#### 🎯 Best Overall Balance")
                st.warning("**EfficientNetB0** - High accuracy + efficiency")
                
            # Model rankings
            st.markdown("### 🏅 Model Rankings")
            
            # Calculate overall scores
            rankings = []
            for model_name, metrics in MODEL_METRICS.items():
                # Weighted score: accuracy (40%), speed (30%), size (30%)
                score = (metrics['accuracy'] * 0.4 + 
                        (100 - metrics['inference_time']) * 0.3 + 
                        (50 - metrics['model_size']) * 0.3)
                rankings.append({'Model': model_name, 'Overall Score': round(score, 1)})
            
            rankings_df = pd.DataFrame(rankings).sort_values('Overall Score', ascending=False)
            st.dataframe(rankings_df, use_container_width=True)
        
        elif page == "🔍 Image Classification":
            st.markdown("## 🔍 Image Classification")
            
            # Instructions
            st.info("""
            📋 **Instructions:**
            1. Upload a satellite or space object image (PNG, JPG, or JPEG)
            2. Select one or more models for classification
            3. Click 'Classify Image' to get predictions
            4. View results, confidence scores, and recommendations
            """)
            
            uploaded_file = st.file_uploader(
                "Upload a satellite image", 
                type=['png', 'jpg', 'jpeg'],
                help="Upload an image of a satellite or space object for classification"
            )
            
            if uploaded_file is not None:
                try:
                    # Display uploaded image
                    image = Image.open(uploaded_file)
                    
                    col1, col2 = st.columns([1, 2])
                    
                    with col1:
                        st.image(image, caption="Uploaded Image", use_container_width=True)
                    
                    with col2:
                        st.markdown("### Image Details")
                        st.write(f"**Filename:** {uploaded_file.name}")
                        st.write(f"**Size:** {image.size}")
                        st.write(f"**Mode:** {image.mode}")
                        st.write(f"**File Size:** {len(uploaded_file.getvalue())} bytes")
                    
                    # Model selection
                    st.markdown("### Select Models for Classification")
                    selected_models = st.multiselect(
                        "Choose models to run predictions with:",
                        list(MODEL_CONFIGS.keys()),
                        default=["EfficientNetB0"],  # Start with just one model to avoid timeouts
                        help="Select one or more models. More models = longer processing time."
                    )
                    
                    if st.button("🚀 Classify Image", type="primary"):
                        if not selected_models:
                            st.warning("Please select at least one model.")
                            return
                        
                        # Preprocess image
                        processed_image = preprocess_image(image)
                        if processed_image is None:
                            st.error("Failed to preprocess image")
                            return
                        
                        # Store predictions
                        predictions = {}
                        
                        # Create progress bar
                        progress_bar = st.progress(0)
                        status_text = st.empty()
                        
                        # Make predictions with selected models
                        for i, model_name in enumerate(selected_models):
                            try:
                                status_text.text(f'Loading {model_name}... ({i+1}/{len(selected_models)})')
                                model = load_model(model_name)
                                
                                if model:
                                    status_text.text(f'Predicting with {model_name}... ({i+1}/{len(selected_models)})')
                                    pred = predict_with_model(model, processed_image, model_name)
                                    if pred:
                                        predictions[model_name] = pred
                                    else:
                                        st.warning(f"Failed to get prediction from {model_name}")
                                else:
                                    st.warning(f"Failed to load {model_name}")
                                
                            except Exception as e:
                                st.error(f"Error processing {model_name}: {str(e)}")
                                logger.error(f"Error processing {model_name}: {str(e)}")
                            
                            progress_bar.progress((i + 1) / len(selected_models))
                        
                        status_text.empty()
                        progress_bar.empty()
                        
                        # Display results
                        if predictions:
                            # Get recommendation
                            recommended_model = recommend_best_model(predictions)
                            
                            st.markdown("### 🎯 Prediction Results")
                            
                            # Show recommendation
                            st.markdown(f"""
                            <div class="prediction-box">
                                <h3>🏆 Recommended Model: {recommended_model}</h3>
                                <p>Based on confidence and model performance</p>
                            </div>
                            """, unsafe_allow_html=True)
                            
                            # Results table
                            results_data = []
                            for model_name, pred in predictions.items():
                                if pred:
                                    results_data.append({
                                        'Model': model_name,
                                        'Predicted Class': pred['class_name'],
                                        'Confidence (%)': f"{pred['confidence']:.1f}%",
                                        'Inference Time (ms)': f"{pred['inference_time']:.1f}",
                                        'Recommended': '🏆' if model_name == recommended_model else ''
                                    })
                            
                            if results_data:
                                df_results = pd.DataFrame(results_data)
                                st.dataframe(df_results, use_container_width=True)
                                
                                # Confidence comparison
                                if len(predictions) > 1:
                                    st.markdown("### 📊 Confidence Comparison")
                                    confidences = [pred['confidence'] for pred in predictions.values() if pred]
                                    model_names = [name for name, pred in predictions.items() if pred]
                                    
                                    try:
                                        fig_conf = go.Figure()
                                        fig_conf.add_trace(go.Bar(
                                            x=model_names,
                                            y=confidences,
                                            marker_color=['gold' if name == recommended_model else 'lightblue' 
                                                        for name in model_names]
                                        ))
                                        fig_conf.update_layout(
                                            title="Prediction Confidence by Model",
                                            xaxis_title="Models",
                                            yaxis_title="Confidence (%)",
                                            height=400
                                        )
                                        st.plotly_chart(fig_conf, use_container_width=True)
                                    except Exception as e:
                                        st.warning(f"Could not create confidence chart: {str(e)}")
                                
                                # Probability distribution for recommended model
                                if recommended_model in predictions and predictions[recommended_model]:
                                    try:
                                        st.markdown(f"### 🔍 Detailed Probabilities - {recommended_model}")
                                        probs = predictions[recommended_model]['probabilities']
                                        prob_df = pd.DataFrame({
                                            'Class': [CLASS_NAMES[i] for i in range(len(probs))],
                                            'Probability': probs * 100
                                        }).sort_values('Probability', ascending=False)
                                        
                                        fig_prob = px.bar(
                                            prob_df.head(5), 
                                            x='Probability', 
                                            y='Class',
                                            orientation='h',
                                            title=f"Top 5 Class Probabilities - {recommended_model}",
                                            color='Probability',
                                            color_continuous_scale='viridis'
                                        )
                                        st.plotly_chart(fig_prob, use_container_width=True)
                                    except Exception as e:
                                        st.warning(f"Could not create probability chart: {str(e)}")
                        else:
                            st.error("No successful predictions were made. Please try again with different models.")
                
                except Exception as e:
                    st.error(f"Error processing uploaded image: {str(e)}")
                    logger.error(f"Error processing uploaded image: {str(e)}")
        
        elif page == "📈 Performance Analytics":
            st.markdown("## 📈 Performance Analytics")
            
            # Performance overview
            col1, col2, col3, col4 = st.columns(4)
            
            with col1:
                st.metric("Best Accuracy", "96.4%", "EfficientNetB0")
            with col2:
                st.metric("Fastest Inference", "18ms", "MobileNetV2")
            with col3:
                st.metric("Smallest Model", "8.7MB", "MobileNetV2")
            with col4:
                st.metric("Total Classes", "11", "Satellites + Debris")
            
            # Detailed analytics
            tab1, tab2, tab3, tab4 = st.tabs(["Accuracy Analysis", "Efficiency Metrics", "Model Comparison", "Confusion Matrix"])
            
            with tab1:
                try:
                    # Accuracy breakdown
                    models = list(MODEL_METRICS.keys())
                    metrics_list = ['accuracy', 'precision', 'recall', 'f1_score']
                    
                    for metric in metrics_list:
                        values = [MODEL_METRICS[model][metric] for model in models]
                        fig = go.Figure()
                        fig.add_trace(go.Bar(
                            x=models, 
                            y=values, 
                            name=metric.title(),
                            marker_color='lightblue',
                            text=[f'{v:.1f}%' for v in values],
                            textposition='auto'
                        ))
                        fig.update_layout(
                            title=f"{metric.title()} Comparison", 
                            height=300,
                            yaxis_title=f"{metric.title()} (%)"
                        )
                        st.plotly_chart(fig, use_container_width=True)
                except Exception as e:
                    st.error(f"Error creating accuracy charts: {str(e)}")
            
            with tab2:
                try:
                    # Efficiency metrics
                    col1, col2 = st.columns(2)
                    
                    with col1:
                        # Inference time
                        times = [MODEL_METRICS[model]["inference_time"] for model in models]
                        fig_time = go.Figure()
                                                fig_time.add_trace(go.Bar(
                            x=models, 
                            y=times,
                            name="Inference Time",
                            marker_color='orange',
                            text=[f'{t:.1f} ms' for t in times],
                            textposition='auto'
                        ))
                        fig_time.update_layout(
                            title="Inference Time per Model",
                            yaxis_title="Time (ms)",
                            height=300
                        )
                        st.plotly_chart(fig_time, use_container_width=True)

                    with col2:
                        # Model sizes
                        sizes = [MODEL_METRICS[model]["model_size"] for model in models]
                        fig_size = go.Figure()
                        fig_size.add_trace(go.Bar(
                            x=models,
                            y=sizes,
                            name="Model Size",
                            marker_color='green',
                            text=[f'{s:.1f} MB' for s in sizes],
                            textposition='auto'
                        ))
                        fig_size.update_layout(
                            title="Model Size per Model",
                            yaxis_title="Size (MB)",
                            height=300
                        )
                        st.plotly_chart(fig_size, use_container_width=True)

                except Exception as e:
                    st.error(f"Error displaying efficiency metrics: {str(e)}")

            with tab3:
                # Reuse full comparison dashboard
                comp_fig = create_metrics_comparison()
                if comp_fig:
                    st.plotly_chart(comp_fig, use_container_width=True)

            with tab4:
                # Display the confusion matrix
                cm_fig = create_confusion_matrix_heatmap()
                if cm_fig:
                    st.plotly_chart(cm_fig, use_container_width=True)

        elif page == "ℹ️ About Models":
            st.markdown("## ℹ️ Model Details and Use Cases")

            for model_name, config in MODEL_CONFIGS.items():
                with st.expander(f"🔍 {model_name}"):
                    st.markdown(f"<div class='model-card'><h4>{model_name}</h4>", unsafe_allow_html=True)
                    st.markdown(f"**Description:** {config['description']}")
                    st.markdown(f"**Input Shape:** {config['input_shape']}")
                    st.markdown("**Strengths:**")
                    for s in config['strengths']:
                        st.markdown(f"• {s}")
                    st.markdown("**Best For:**")
                    for use in config['best_for']:
                        st.markdown(f"• {use}")
                    st.markdown("</div>", unsafe_allow_html=True)

    except Exception as e:
        st.error(f"An unexpected error occurred: {str(e)}")
        logger.error(f"Main app error: {str(e)}")


# Run the app
if __name__ == "__main__":
    main()