Initial leaderboard setup

.ipynb_checkpoints/app-checkpoint.py

@@ -9,15 +9,40 @@ import io
 import os
 import base64
 
-# Initialize session state
-if 'active_tab' not in st.session_state:
-    st.session_state.active_tab = "📈 Leaderboard"
+# Page config
+st.set_page_config(
+    page_title="Nexar Driving Leaderboard",
+    page_icon="nexar_logo.png",
+    layout="wide"
+)
 
-if 'toggle_states' not in st.session_state:
-    st.session_state.toggle_states = {}
+# Custom styling
+st.markdown("""
+    <style>
+    .main { padding: 2rem; }
+    .stTabs [data-baseweb="tab-list"] { gap: 8px; }
+    .stTabs [data-baseweb="tab"] {
+        padding: 8px 16px;
+        border-radius: 4px;
+    }
+    .metric-card {
+        background-color: #f8f9fa;
+        padding: 20px;
+        border-radius: 10px;
+        box-shadow: 0 2px 4px rgba(0,0,0,0.1);
+    }
+    </style>
+""", unsafe_allow_html=True)
+
+# Header
+col1, col2 = st.columns([0.16, 0.84])
+with col1:
+    st.image("nexar_logo.png", width=600)
+with col2:
+    st.title("Driving Leaderboard")
 
 # Data loading function
-@st.cache_data
+@st.cache_data(experimental_allow_widgets=True)
 def load_data(directory='results', labels_filename='Labels.csv'):
     labels_path = os.path.join(directory, labels_filename)
     df_labels = pd.read_csv(labels_path)
@@ -33,7 +58,15 @@ def load_data(directory='results', labels_filename='Labels.csv'):
     model_comparison = ModelComparison(evaluators)
     return model_comparison
 
-# Helper functions for styling
+# Initialize session state
+if 'model_comparison' not in st.session_state:
+    st.session_state.model_comparison = load_data()
+    st.session_state.leaderboard_df = st.session_state.model_comparison.transform_to_leaderboard()
+    st.session_state.combined_df = st.session_state.model_comparison.combined_df
+
+# Create tabs
+tab1, tab2, tab3, tab4 = st.tabs(["📈 Leaderboard", "📊 Class Performance", "🔍 Detailed Metrics", "⚖️ Model Comparison"])
+
 def style_dataframe(df, highlight_first_column=True, show_progress_bars=True):
     numeric_cols = df.select_dtypes(include=['float64']).columns
     
@@ -77,70 +110,8 @@ def style_dataframe(df, highlight_first_column=True, show_progress_bars=True):
         ])
     return styled
 
-# Toggle state management
-def get_toggle_state(model_name):
-    key = f"toggle_{model_name}"
-    if key not in st.session_state.toggle_states:
-        st.session_state.toggle_states[key] = True
-    return st.session_state.toggle_states[key]
-
-def set_toggle_state(model_name, value):
-    key = f"toggle_{model_name}"
-    st.session_state.toggle_states[key] = value
-
-# Page configuration
-st.set_page_config(
-    page_title="Nexar Driving Leaderboard",
-    page_icon="nexar_logo.png",
-    layout="wide"
-)
-
-# Custom styling
-st.markdown("""
-    <style>
-    .main { padding: 2rem; }
-    .stTabs [data-baseweb="tab-list"] { gap: 8px; }
-    .stTabs [data-baseweb="tab"] {
-        padding: 8px 16px;
-        border-radius: 4px;
-    }
-    .metric-card {
-        background-color: #f8f9fa;
-        padding: 20px;
-        border-radius: 10px;
-        box-shadow: 0 2px 4px rgba(0,0,0,0.1);
-    }
-    </style>
-""", unsafe_allow_html=True)
-
-# Header
-col1, col2 = st.columns([0.16, 0.84])
-with col1:
-    st.image("nexar_logo.png", width=600)
-with col2:
-    st.title("Driving Leaderboard")
-
-# Initialize data in session state
-if 'model_comparison' not in st.session_state:
-    st.session_state.model_comparison = load_data()
-    st.session_state.leaderboard_df = st.session_state.model_comparison.transform_to_leaderboard()
-    st.session_state.combined_df = st.session_state.model_comparison.combined_df
-
-# Tab callback
-def handle_tab_change(tab_name):
-    st.session_state.active_tab = tab_name
-
-# Define tab names
-tab_names = ["📈 Leaderboard", "📊 Class Performance", "🔍 Detailed Metrics", "⚖️ Model Comparison"]
-
-# Create tabs
-selected_tab = st.radio("", tab_names, key="tab_selector", 
-                       horizontal=True, label_visibility="collapsed",
-                       index=tab_names.index(st.session_state.active_tab))
-handle_tab_change(selected_tab)
-
-# Content based on selected tab
-if st.session_state.active_tab == "📈 Leaderboard":
+# Tab 1: Leaderboard
+with tab1:
     st.subheader("Model Performance Leaderboard")
     
     sort_col = st.selectbox(
@@ -150,7 +121,11 @@ if st.session_state.active_tab == "📈 Leaderboard":
     )
     
     sorted_df = st.session_state.leaderboard_df.sort_values(by=sort_col, ascending=False)
-    st.dataframe(style_dataframe(sorted_df), use_container_width=True)
+    
+    st.dataframe(
+        style_dataframe(sorted_df),
+        use_container_width=True,
+    )
     
     metrics = ['F1 Score', 'Precision', 'Recall']
     selected_metric = st.selectbox("Select Metric for Category Analysis:", metrics)
@@ -176,10 +151,11 @@ if st.session_state.active_tab == "📈 Leaderboard":
     
     st.plotly_chart(fig, use_container_width=True)
 
-elif st.session_state.active_tab == "📊 Class Performance":
+# Tab 2: Class Performance
+with tab2:
     st.subheader("Class-level Performance")
     categories = st.session_state.combined_df['Category'].unique()
-    metrics = ['F1 Score', 'Precision', 'Recall']
+
     col1, col2, col3 = st.columns(3)
     with col1:
         selected_category = st.selectbox(
@@ -194,26 +170,23 @@ elif st.session_state.active_tab == "📊 Class Performance":
             key='class_metric'
         )
     with col3:
-        all_models = sorted(st.session_state.combined_df['Model'].unique())
         selected_models = st.multiselect(
             "Select Models:",
-            all_models,
-            default=all_models,
-            key='selected_models'
+            st.session_state.combined_df['Model'].unique(),
+            default=st.session_state.combined_df['Model'].unique()
         )
-
-    # Create consistent color mapping
+    
+    # Create a consistent color mapping for all models
     plotly_colors = ['#636EFA', '#EF553B', '#00CC96', '#AB63FA', '#FFA15A', '#19D3F3', '#FF6692', '#B6E880', '#FF97FF', '#FECB52']
-    model_colors = {model: plotly_colors[i % len(plotly_colors)] for i, model in enumerate(all_models)}
+    model_colors = {model: plotly_colors[i % len(plotly_colors)] for i, model in enumerate(sorted(st.session_state.combined_df['Model'].unique()))}
     
-    # Filter data
     class_data = st.session_state.combined_df[
         (st.session_state.combined_df['Category'] == selected_category) &
         (~st.session_state.combined_df['Class'].str.contains('Overall')) &
         (st.session_state.combined_df['Model'].isin(selected_models))
     ]
     
-    # Bar chart
+    # Bar chart with consistent colors
     fig = px.bar(
         class_data,
         x='Class',
@@ -225,11 +198,13 @@ elif st.session_state.active_tab == "📊 Class Performance":
     )
     st.plotly_chart(fig, use_container_width=True)
     
-    # Model toggles
-    st.markdown("### Model Visibility Controls")
+    # Calculate how many columns we need (aim for about 4-5 models per row)
     models_per_row = 4
     num_rows = (len(selected_models) + models_per_row - 1) // models_per_row
     
+    st.markdown("### Select Models to Display:")
+    
+    # Create toggles for models using st.columns
     for row in range(num_rows):
         cols = st.columns(models_per_row)
         for col_idx in range(models_per_row):
@@ -237,44 +212,50 @@ elif st.session_state.active_tab == "📊 Class Performance":
             if model_idx < len(selected_models):
                 model = selected_models[model_idx]
                 container = cols[col_idx].container()
+                
+                # Get the consistent color for this model
                 color = model_colors[model]
                 
-                # Create colored indicator
+                # Initialize toggle state if needed
+                toggle_key = f"toggle_{model}"
+                if toggle_key not in st.session_state:
+                    st.session_state[toggle_key] = True
+                
+                # Create colored legend item with HTML
                 container.markdown(
                     f"""
-                    <div style='display: flex; align-items: center; margin-bottom: -40px;'>
+                    <div style='display: flex; align-items: center; margin-bottom: -40px; pointer-events: none;'>
                         <span style='display: inline-block; width: 12px; height: 12px; background-color: {color}; border-radius: 50%; margin-right: 8px;'></span>
                     </div>
                     """, 
                     unsafe_allow_html=True
                 )
                 
-                # Toggle checkbox
-                value = container.checkbox(
-                    f"    {model}",
-                    value=get_toggle_state(model),
-                    key=f"vis_{model}",
-                    on_change=set_toggle_state,
-                    args=(model, not get_toggle_state(model))
+                # Create the checkbox without reassigning to session state
+                container.checkbox(
+                    f"    {model}",  # Add some spacing to account for the circle
+                    value=st.session_state[toggle_key],
+                    key=toggle_key  # Use toggle_key directly as the key
                 )
     
-    # Precision-Recall plots
-    st.markdown("### Precision-Recall Analysis by Class")
+    # Individual Precision-Recall plots for each class
     unique_classes = class_data['Class'].unique()
     num_classes = len(unique_classes)
-    plots_per_row = 3
-    num_plot_rows = (num_classes + plots_per_row - 1) // plots_per_row
     
-    for row in range(num_plot_rows):
-        cols = st.columns(plots_per_row)
-        for col_idx in range(plots_per_row):
-            class_idx = row * plots_per_row + col_idx
+    # Calculate number of rows needed (3 plots per row)
+    num_rows = (num_classes + 2) // 3  # Using ceiling division
+    
+    # Create plots row by row
+    for row in range(num_rows):
+        cols = st.columns(3)
+        for col_idx in range(3):
+            class_idx = row * 3 + col_idx
             if class_idx < num_classes:
                 current_class = unique_classes[class_idx]
                 
-                # Get visible models
+                # Filter data based on visible models
                 visible_models = [model for model in selected_models 
-                                if get_toggle_state(model)]
+                                if st.session_state[f"toggle_{model}"]]
                 
                 class_specific_data = class_data[
                     (class_data['Class'] == current_class) & 
@@ -288,16 +269,18 @@ elif st.session_state.active_tab == "📊 Class Performance":
                     color='Model',
                     title=f'Precision vs Recall: {current_class}',
                     height=300,
-                    color_discrete_map=model_colors
+                    color_discrete_map=model_colors  # Use consistent colors
                 )
                 
+                # Update layout for better visibility
                 fig.update_layout(
                     xaxis_range=[0, 1],
                     yaxis_range=[0, 1],
                     margin=dict(l=40, r=40, t=40, b=40),
-                    showlegend=False
+                    showlegend=False  # Hide individual legends
                 )
                 
+                # Add diagonal reference line
                 fig.add_trace(
                     go.Scatter(
                         x=[0, 1],
@@ -310,53 +293,74 @@ elif st.session_state.active_tab == "📊 Class Performance":
                 
                 cols[col_idx].plotly_chart(fig, use_container_width=True)
 
-elif st.session_state.active_tab == "🔍 Detailed Metrics":
+# Tab 3: Detailed Metrics
+with tab3:
     st.subheader("Detailed Metrics Analysis")
     
     selected_model = st.selectbox(
         "Select Model for Detailed Analysis:",
-        st.session_state.combined_df['Model'].unique(),
-        key='detailed_model'
+        st.session_state.combined_df['Model'].unique()
     )
     
     model_data = st.session_state.combined_df[
         st.session_state.combined_df['Model'] == selected_model
     ]
     
+    # Create metrics tables
     st.markdown("### Performance Metrics by Category")
+    
+    # Get unique categories and relevant classes for each category
     categories = model_data['Category'].unique()
     metrics = ['F1 Score', 'Precision', 'Recall']
     
+    # Process data for each category
     for category in categories:
         st.markdown(f"#### {category}")
+        
+        # Filter data for this category
         category_data = model_data[model_data['Category'] == category].copy()
         
-        # Get classes excluding Overall
+        # Create a clean table for this category
+        category_metrics = pd.DataFrame()
+        
+        # Get classes for this category (excluding 'Overall' prefix)
         classes = category_data[~category_data['Class'].str.contains('Overall')]['Class'].unique()
+        
+        # Add the overall metric for this category
         overall_data = category_data[category_data['Class'].str.contains('Overall')]
         
-        # Create metrics DataFrame
+        # Initialize the DataFrame with classes as index
         category_metrics = pd.DataFrame(index=classes)
+        
+        # Add metrics columns
         for metric in metrics:
+            # Add class-specific metrics
             class_metrics = {}
             for class_name in classes:
                 class_data = category_data[category_data['Class'] == class_name]
                 if not class_data.empty:
                     class_metrics[class_name] = class_data[metric].iloc[0]
+            
             category_metrics[metric] = pd.Series(class_metrics)
         
-        # Add overall metrics
+        # Add overall metrics as a separate row
         if not overall_data.empty:
             overall_row = pd.DataFrame({
                 metric: [overall_data[metric].iloc[0]] for metric in metrics
             }, index=['Overall'])
             category_metrics = pd.concat([overall_row, category_metrics])
         
-        st.dataframe(style_dataframe(category_metrics.round(4)), use_container_width=True)
+        # Display the table
+        styled_metrics = style_dataframe(category_metrics.round(4))
+        st.dataframe(styled_metrics, use_container_width=True)
+        
+        # Add spacing between categories
         st.markdown("---")
     
     # Export functionality
     st.markdown("### Export Data")
+    
+    # Prepare export data
     export_data = pd.DataFrame()
     for category in categories:
         category_data = model_data[model_data['Category'] == category].copy()
@@ -368,6 +372,7 @@ elif st.session_state.active_tab == "🔍 Detailed Metrics":
         ).round(4)
         export_data = pd.concat([export_data, category_metrics])
     
+    # Create download button
     csv = export_data.to_csv().encode()
     st.download_button(
         "Download Detailed Metrics",
@@ -377,25 +382,31 @@ elif st.session_state.active_tab == "🔍 Detailed Metrics":
         key='download-csv'
     )
 
-elif st.session_state.active_tab == "⚖️ Model Comparison":
+# Tab 4: Model Comparison
+with tab4:
     st.header("Model Comparison Analysis")
     
+    # Create two columns for model selection
     col1, col2 = st.columns(2)
+    
+    # Model selection dropdown menus
     with col1:
         model1 = st.selectbox(
             "Select First Model:",
             st.session_state.combined_df['Model'].unique(),
-            key='compare_model1'
+            key='model1'
         )
     
     with col2:
+        # Filter out the first selected model from options
         available_models = [m for m in st.session_state.combined_df['Model'].unique() if m != model1]
         model2 = st.selectbox(
             "Select Second Model:",
             available_models,
-            key='compare_model2'
+            key='model2'
         )
     
+    # Category selection
     selected_category = st.selectbox(
         "Select Category for Comparison:",
         st.session_state.combined_df['Category'].unique(),
@@ -412,19 +423,26 @@ elif st.session_state.active_tab == "⚖️ Model Comparison":
         (st.session_state.combined_df['Model'] == model2) &
         (st.session_state.combined_df['Category'] == selected_category)
     ]
+
+    # Define metrics list
+    metrics = ['F1 Score', 'Precision', 'Recall']
     
+    # Create comparison tables section
     st.subheader("Detailed Metrics Comparison")
-    metrics = ['F1 Score', 'Precision', 'Recall']
     
+    # Create a table for each metric
     for metric in metrics:
         st.markdown(f"#### {metric} Comparison")
-        metric_data = []
         
+        # Prepare data for the metric table
+        metric_data = []
         for class_name in model1_data['Class'].unique():
-            m1_value = model1_data[model1_data['Class'] == class_name][metric].iloc[0] * 100
-            m2_value = model2_data[model2_data['Class'] == class_name][metric].iloc[0] * 100
+            # Get values for both models
+            m1_value = model1_data[model1_data['Class'] == class_name][metric].iloc[0]
+            m2_value = model2_data[model2_data['Class'] == class_name][metric].iloc[0]
             diff = m1_value - m2_value
             
+            # Add to comparison data
             metric_data.append({
                 'Class': class_name,
                 model1: m1_value,
@@ -432,46 +450,92 @@ elif st.session_state.active_tab == "⚖️ Model Comparison":
                 'Difference': diff
             })
         
+        # Create DataFrame for the metric
         metric_df = pd.DataFrame(metric_data)
-        st.dataframe(style_dataframe(metric_df), use_container_width=True)
+        
+        # Style the table
+        def style_metric_table(df):
+            return df.style\
+                .format({
+                    model1: '{:.2f}%',
+                    model2: '{:.2f}%',
+                    'Difference': '{:+.2f}%'
+                })\
+                .background_gradient(
+                    cmap='RdYlGn',
+                    subset=['Difference'],
+                    vmin=-10,
+                    vmax=10
+                )\
+                .set_properties(**{
+                    'text-align': 'center',
+                    'padding': '10px',
+                    'border': '1px solid #dee2e6'
+                })\
+                .set_table_styles([
+                    {'selector': 'th', 'props': [
+                        ('background-color', '#4a90e2'),
+                        ('color', 'white'),
+                        ('font-weight', 'bold'),
+                        ('text-align', 'center'),
+                        ('padding', '10px')
+                    ]}
+                ])
+        
+        # Display the styled table
+        
+        st.dataframe(
+            style_dataframe(metric_df),
+            use_container_width=True,
+        )
+        # Add visual separator
         st.markdown("---")
     
-    # Visual comparison
+    # Visualizations section
     st.subheader("Visual Performance Analysis")
+    
+    # Metric selector for bar chart
     selected_metric = st.selectbox(
         "Select Metric for Comparison:",
         metrics,
-        key='visual_compare_metric'
+        key='compare_metric'
     )
     
-    # Prepare data for visualization
+    # Prepare data for bar chart
     comparison_data = pd.DataFrame()
+    
+    # Get data for both models
     for idx, (model_name, model_data) in enumerate([(model1, model1_data), (model2, model2_data)]):
+        # Filter out Overall classes and select relevant columns
         model_metrics = model_data[~model_data['Class'].str.contains('Overall', na=False)][['Class', selected_metric]]
         model_metrics = model_metrics.rename(columns={selected_metric: model_name})
         
+        # Merge with existing data or create new DataFrame
         if idx == 0:
             comparison_data = model_metrics
         else:
             comparison_data = comparison_data.merge(model_metrics, on='Class', how='outer')
     
-    # Bar chart
+    # Create bar chart
     fig_bar = go.Figure()
     
+    # Add bars for first model
     fig_bar.add_trace(go.Bar(
         name=model1,
         x=comparison_data['Class'],
-        y=comparison_data[model1] * 100,
+        y=comparison_data[model1],
         marker_color='rgb(55, 83, 109)'
     ))
     
+    # Add bars for second model
     fig_bar.add_trace(go.Bar(
         name=model2,
         x=comparison_data['Class'],
-        y=comparison_data[model2] * 100,
+        y=comparison_data[model2],
         marker_color='rgb(26, 118, 255)'
     ))
     
+    # Update bar chart layout
     fig_bar.update_layout(
         title=f"{selected_metric} Comparison by Class",
         xaxis_title="Class",
@@ -488,19 +552,23 @@ elif st.session_state.active_tab == "⚖️ Model Comparison":
         )
     )
     
+    # Display bar chart
     st.plotly_chart(fig_bar, use_container_width=True)
     
-    # Precision-Recall Analysis
+    # Create Precision-Recall scatter plot
     st.markdown("#### Precision-Recall Analysis")
     
+    # Filter data for scatter plot
     model1_scatter = model1_data[~model1_data['Class'].str.contains('Overall', na=False)]
     model2_scatter = model2_data[~model2_data['Class'].str.contains('Overall', na=False)]
     
+    # Create scatter plot
     fig_scatter = go.Figure()
     
+    # Add scatter points for first model
     fig_scatter.add_trace(go.Scatter(
-        x=model1_scatter['Precision'] * 100,
-        y=model1_scatter['Recall'] * 100,
+        x=model1_scatter['Precision']*100,
+        y=model1_scatter['Recall']*100,
         mode='markers+text',
         name=model1,
         text=model1_scatter['Class'],
@@ -508,9 +576,10 @@ elif st.session_state.active_tab == "⚖️ Model Comparison":
         marker=dict(size=10)
     ))
     
+    # Add scatter points for second model
     fig_scatter.add_trace(go.Scatter(
-        x=model2_scatter['Precision'] * 100,
-        y=model2_scatter['Recall'] * 100,
+        x=model2_scatter['Precision']*100,
+        y=model2_scatter['Recall']*100,
         mode='markers+text',
         name=model2,
         text=model2_scatter['Class'],
@@ -518,6 +587,7 @@ elif st.session_state.active_tab == "⚖️ Model Comparison":
         marker=dict(size=10)
     ))
     
+    # Add reference line
     fig_scatter.add_trace(go.Scatter(
         x=[0, 100],
         y=[0, 100],
@@ -526,6 +596,7 @@ elif st.session_state.active_tab == "⚖️ Model Comparison":
         showlegend=False
     ))
     
+    # Update scatter plot layout
     fig_scatter.update_layout(
         title="Precision vs Recall Analysis by Class",
         xaxis_title="Precision (%)",
@@ -542,8 +613,10 @@ elif st.session_state.active_tab == "⚖️ Model Comparison":
         )
     )
     
+    # Display scatter plot
     st.plotly_chart(fig_scatter, use_container_width=True)
 
+
 # Footer
 st.markdown("---")
 st.markdown("Dashboard created for model evaluation and comparison")

app.py

@@ -9,15 +9,40 @@ import io
 import os
 import base64
 
-# Initialize session state
-if 'active_tab' not in st.session_state:
-    st.session_state.active_tab = "📈 Leaderboard"
+# Page config
+st.set_page_config(
+    page_title="Nexar Driving Leaderboard",
+    page_icon="nexar_logo.png",
+    layout="wide"
+)
 
-if 'toggle_states' not in st.session_state:
-    st.session_state.toggle_states = {}
+# Custom styling
+st.markdown("""
+    <style>
+    .main { padding: 2rem; }
+    .stTabs [data-baseweb="tab-list"] { gap: 8px; }
+    .stTabs [data-baseweb="tab"] {
+        padding: 8px 16px;
+        border-radius: 4px;
+    }
+    .metric-card {
+        background-color: #f8f9fa;
+        padding: 20px;
+        border-radius: 10px;
+        box-shadow: 0 2px 4px rgba(0,0,0,0.1);
+    }
+    </style>
+""", unsafe_allow_html=True)
+
+# Header
+col1, col2 = st.columns([0.16, 0.84])
+with col1:
+    st.image("nexar_logo.png", width=600)
+with col2:
+    st.title("Driving Leaderboard")
 
 # Data loading function
-@st.cache_data
+@st.cache_data(experimental_allow_widgets=True)
 def load_data(directory='results', labels_filename='Labels.csv'):
     labels_path = os.path.join(directory, labels_filename)
     df_labels = pd.read_csv(labels_path)
@@ -33,7 +58,15 @@ def load_data(directory='results', labels_filename='Labels.csv'):
     model_comparison = ModelComparison(evaluators)
     return model_comparison
 
-# Helper functions for styling
+# Initialize session state
+if 'model_comparison' not in st.session_state:
+    st.session_state.model_comparison = load_data()
+    st.session_state.leaderboard_df = st.session_state.model_comparison.transform_to_leaderboard()
+    st.session_state.combined_df = st.session_state.model_comparison.combined_df
+
+# Create tabs
+tab1, tab2, tab3, tab4 = st.tabs(["📈 Leaderboard", "📊 Class Performance", "🔍 Detailed Metrics", "⚖️ Model Comparison"])
+
 def style_dataframe(df, highlight_first_column=True, show_progress_bars=True):
     numeric_cols = df.select_dtypes(include=['float64']).columns
     
@@ -77,70 +110,8 @@ def style_dataframe(df, highlight_first_column=True, show_progress_bars=True):
         ])
     return styled
 
-# Toggle state management
-def get_toggle_state(model_name):
-    key = f"toggle_{model_name}"
-    if key not in st.session_state.toggle_states:
-        st.session_state.toggle_states[key] = True
-    return st.session_state.toggle_states[key]
-
-def set_toggle_state(model_name, value):
-    key = f"toggle_{model_name}"
-    st.session_state.toggle_states[key] = value
-
-# Page configuration
-st.set_page_config(
-    page_title="Nexar Driving Leaderboard",
-    page_icon="nexar_logo.png",
-    layout="wide"
-)
-
-# Custom styling
-st.markdown("""
-    <style>
-    .main { padding: 2rem; }
-    .stTabs [data-baseweb="tab-list"] { gap: 8px; }
-    .stTabs [data-baseweb="tab"] {
-        padding: 8px 16px;
-        border-radius: 4px;
-    }
-    .metric-card {
-        background-color: #f8f9fa;
-        padding: 20px;
-        border-radius: 10px;
-        box-shadow: 0 2px 4px rgba(0,0,0,0.1);
-    }
-    </style>
-""", unsafe_allow_html=True)
-
-# Header
-col1, col2 = st.columns([0.16, 0.84])
-with col1:
-    st.image("nexar_logo.png", width=600)
-with col2:
-    st.title("Driving Leaderboard")
-
-# Initialize data in session state
-if 'model_comparison' not in st.session_state:
-    st.session_state.model_comparison = load_data()
-    st.session_state.leaderboard_df = st.session_state.model_comparison.transform_to_leaderboard()
-    st.session_state.combined_df = st.session_state.model_comparison.combined_df
-
-# Tab callback
-def handle_tab_change(tab_name):
-    st.session_state.active_tab = tab_name
-
-# Define tab names
-tab_names = ["📈 Leaderboard", "📊 Class Performance", "🔍 Detailed Metrics", "⚖️ Model Comparison"]
-
-# Create tabs
-selected_tab = st.radio("", tab_names, key="tab_selector", 
-                       horizontal=True, label_visibility="collapsed",
-                       index=tab_names.index(st.session_state.active_tab))
-handle_tab_change(selected_tab)
-
-# Content based on selected tab
-if st.session_state.active_tab == "📈 Leaderboard":
+# Tab 1: Leaderboard
+with tab1:
     st.subheader("Model Performance Leaderboard")
     
     sort_col = st.selectbox(
@@ -150,7 +121,11 @@ if st.session_state.active_tab == "📈 Leaderboard":
     )
     
     sorted_df = st.session_state.leaderboard_df.sort_values(by=sort_col, ascending=False)
-    st.dataframe(style_dataframe(sorted_df), use_container_width=True)
+    
+    st.dataframe(
+        style_dataframe(sorted_df),
+        use_container_width=True,
+    )
     
     metrics = ['F1 Score', 'Precision', 'Recall']
     selected_metric = st.selectbox("Select Metric for Category Analysis:", metrics)
@@ -176,10 +151,11 @@ if st.session_state.active_tab == "📈 Leaderboard":
     
     st.plotly_chart(fig, use_container_width=True)
 
-elif st.session_state.active_tab == "📊 Class Performance":
+# Tab 2: Class Performance
+with tab2:
     st.subheader("Class-level Performance")
     categories = st.session_state.combined_df['Category'].unique()
-    metrics = ['F1 Score', 'Precision', 'Recall']
+
     col1, col2, col3 = st.columns(3)
     with col1:
         selected_category = st.selectbox(
@@ -194,26 +170,23 @@ elif st.session_state.active_tab == "📊 Class Performance":
             key='class_metric'
         )
     with col3:
-        all_models = sorted(st.session_state.combined_df['Model'].unique())
         selected_models = st.multiselect(
             "Select Models:",
-            all_models,
-            default=all_models,
-            key='selected_models'
+            st.session_state.combined_df['Model'].unique(),
+            default=st.session_state.combined_df['Model'].unique()
         )
-
-    # Create consistent color mapping
+    
+    # Create a consistent color mapping for all models
     plotly_colors = ['#636EFA', '#EF553B', '#00CC96', '#AB63FA', '#FFA15A', '#19D3F3', '#FF6692', '#B6E880', '#FF97FF', '#FECB52']
-    model_colors = {model: plotly_colors[i % len(plotly_colors)] for i, model in enumerate(all_models)}
+    model_colors = {model: plotly_colors[i % len(plotly_colors)] for i, model in enumerate(sorted(st.session_state.combined_df['Model'].unique()))}
     
-    # Filter data
     class_data = st.session_state.combined_df[
         (st.session_state.combined_df['Category'] == selected_category) &
         (~st.session_state.combined_df['Class'].str.contains('Overall')) &
         (st.session_state.combined_df['Model'].isin(selected_models))
     ]
     
-    # Bar chart
+    # Bar chart with consistent colors
     fig = px.bar(
         class_data,
         x='Class',
@@ -225,11 +198,13 @@ elif st.session_state.active_tab == "📊 Class Performance":
     )
     st.plotly_chart(fig, use_container_width=True)
     
-    # Model toggles
-    st.markdown("### Model Visibility Controls")
+    # Calculate how many columns we need (aim for about 4-5 models per row)
     models_per_row = 4
     num_rows = (len(selected_models) + models_per_row - 1) // models_per_row
     
+    st.markdown("### Select Models to Display:")
+    
+    # Create toggles for models using st.columns
     for row in range(num_rows):
         cols = st.columns(models_per_row)
         for col_idx in range(models_per_row):
@@ -237,44 +212,50 @@ elif st.session_state.active_tab == "📊 Class Performance":
             if model_idx < len(selected_models):
                 model = selected_models[model_idx]
                 container = cols[col_idx].container()
+                
+                # Get the consistent color for this model
                 color = model_colors[model]
                 
-                # Create colored indicator
+                # Initialize toggle state if needed
+                toggle_key = f"toggle_{model}"
+                if toggle_key not in st.session_state:
+                    st.session_state[toggle_key] = True
+                
+                # Create colored legend item with HTML
                 container.markdown(
                     f"""
-                    <div style='display: flex; align-items: center; margin-bottom: -40px;'>
+                    <div style='display: flex; align-items: center; margin-bottom: -40px; pointer-events: none;'>
                         <span style='display: inline-block; width: 12px; height: 12px; background-color: {color}; border-radius: 50%; margin-right: 8px;'></span>
                     </div>
                     """, 
                     unsafe_allow_html=True
                 )
                 
-                # Toggle checkbox
-                value = container.checkbox(
-                    f"    {model}",
-                    value=get_toggle_state(model),
-                    key=f"vis_{model}",
-                    on_change=set_toggle_state,
-                    args=(model, not get_toggle_state(model))
+                # Create the checkbox without reassigning to session state
+                container.checkbox(
+                    f"    {model}",  # Add some spacing to account for the circle
+                    value=st.session_state[toggle_key],
+                    key=toggle_key  # Use toggle_key directly as the key
                 )
     
-    # Precision-Recall plots
-    st.markdown("### Precision-Recall Analysis by Class")
+    # Individual Precision-Recall plots for each class
     unique_classes = class_data['Class'].unique()
     num_classes = len(unique_classes)
-    plots_per_row = 3
-    num_plot_rows = (num_classes + plots_per_row - 1) // plots_per_row
     
-    for row in range(num_plot_rows):
-        cols = st.columns(plots_per_row)
-        for col_idx in range(plots_per_row):
-            class_idx = row * plots_per_row + col_idx
+    # Calculate number of rows needed (3 plots per row)
+    num_rows = (num_classes + 2) // 3  # Using ceiling division
+    
+    # Create plots row by row
+    for row in range(num_rows):
+        cols = st.columns(3)
+        for col_idx in range(3):
+            class_idx = row * 3 + col_idx
             if class_idx < num_classes:
                 current_class = unique_classes[class_idx]
                 
-                # Get visible models
+                # Filter data based on visible models
                 visible_models = [model for model in selected_models 
-                                if get_toggle_state(model)]
+                                if st.session_state[f"toggle_{model}"]]
                 
                 class_specific_data = class_data[
                     (class_data['Class'] == current_class) & 
@@ -288,16 +269,18 @@ elif st.session_state.active_tab == "📊 Class Performance":
                     color='Model',
                     title=f'Precision vs Recall: {current_class}',
                     height=300,
-                    color_discrete_map=model_colors
+                    color_discrete_map=model_colors  # Use consistent colors
                 )
                 
+                # Update layout for better visibility
                 fig.update_layout(
                     xaxis_range=[0, 1],
                     yaxis_range=[0, 1],
                     margin=dict(l=40, r=40, t=40, b=40),
-                    showlegend=False
+                    showlegend=False  # Hide individual legends
                 )
                 
+                # Add diagonal reference line
                 fig.add_trace(
                     go.Scatter(
                         x=[0, 1],
@@ -310,53 +293,74 @@ elif st.session_state.active_tab == "📊 Class Performance":
                 
                 cols[col_idx].plotly_chart(fig, use_container_width=True)
 
-elif st.session_state.active_tab == "🔍 Detailed Metrics":
+# Tab 3: Detailed Metrics
+with tab3:
     st.subheader("Detailed Metrics Analysis")
     
     selected_model = st.selectbox(
         "Select Model for Detailed Analysis:",
-        st.session_state.combined_df['Model'].unique(),
-        key='detailed_model'
+        st.session_state.combined_df['Model'].unique()
     )
     
     model_data = st.session_state.combined_df[
         st.session_state.combined_df['Model'] == selected_model
     ]
     
+    # Create metrics tables
     st.markdown("### Performance Metrics by Category")
+    
+    # Get unique categories and relevant classes for each category
     categories = model_data['Category'].unique()
     metrics = ['F1 Score', 'Precision', 'Recall']
     
+    # Process data for each category
     for category in categories:
         st.markdown(f"#### {category}")
+        
+        # Filter data for this category
         category_data = model_data[model_data['Category'] == category].copy()
         
-        # Get classes excluding Overall
+        # Create a clean table for this category
+        category_metrics = pd.DataFrame()
+        
+        # Get classes for this category (excluding 'Overall' prefix)
         classes = category_data[~category_data['Class'].str.contains('Overall')]['Class'].unique()
+        
+        # Add the overall metric for this category
         overall_data = category_data[category_data['Class'].str.contains('Overall')]
         
-        # Create metrics DataFrame
+        # Initialize the DataFrame with classes as index
         category_metrics = pd.DataFrame(index=classes)
+        
+        # Add metrics columns
         for metric in metrics:
+            # Add class-specific metrics
             class_metrics = {}
             for class_name in classes:
                 class_data = category_data[category_data['Class'] == class_name]
                 if not class_data.empty:
                     class_metrics[class_name] = class_data[metric].iloc[0]
+            
             category_metrics[metric] = pd.Series(class_metrics)
         
-        # Add overall metrics
+        # Add overall metrics as a separate row
         if not overall_data.empty:
             overall_row = pd.DataFrame({
                 metric: [overall_data[metric].iloc[0]] for metric in metrics
             }, index=['Overall'])
             category_metrics = pd.concat([overall_row, category_metrics])
         
-        st.dataframe(style_dataframe(category_metrics.round(4)), use_container_width=True)
+        # Display the table
+        styled_metrics = style_dataframe(category_metrics.round(4))
+        st.dataframe(styled_metrics, use_container_width=True)
+        
+        # Add spacing between categories
         st.markdown("---")
     
     # Export functionality
     st.markdown("### Export Data")
+    
+    # Prepare export data
     export_data = pd.DataFrame()
     for category in categories:
         category_data = model_data[model_data['Category'] == category].copy()
@@ -368,6 +372,7 @@ elif st.session_state.active_tab == "🔍 Detailed Metrics":
         ).round(4)
         export_data = pd.concat([export_data, category_metrics])
     
+    # Create download button
     csv = export_data.to_csv().encode()
     st.download_button(
         "Download Detailed Metrics",
@@ -377,25 +382,31 @@ elif st.session_state.active_tab == "🔍 Detailed Metrics":
         key='download-csv'
     )
 
-elif st.session_state.active_tab == "⚖️ Model Comparison":
+# Tab 4: Model Comparison
+with tab4:
     st.header("Model Comparison Analysis")
     
+    # Create two columns for model selection
     col1, col2 = st.columns(2)
+    
+    # Model selection dropdown menus
     with col1:
         model1 = st.selectbox(
             "Select First Model:",
             st.session_state.combined_df['Model'].unique(),
-            key='compare_model1'
+            key='model1'
         )
     
     with col2:
+        # Filter out the first selected model from options
         available_models = [m for m in st.session_state.combined_df['Model'].unique() if m != model1]
         model2 = st.selectbox(
             "Select Second Model:",
             available_models,
-            key='compare_model2'
+            key='model2'
         )
     
+    # Category selection
     selected_category = st.selectbox(
         "Select Category for Comparison:",
         st.session_state.combined_df['Category'].unique(),
@@ -412,19 +423,26 @@ elif st.session_state.active_tab == "⚖️ Model Comparison":
         (st.session_state.combined_df['Model'] == model2) &
         (st.session_state.combined_df['Category'] == selected_category)
     ]
+
+    # Define metrics list
+    metrics = ['F1 Score', 'Precision', 'Recall']
     
+    # Create comparison tables section
     st.subheader("Detailed Metrics Comparison")
-    metrics = ['F1 Score', 'Precision', 'Recall']
     
+    # Create a table for each metric
     for metric in metrics:
         st.markdown(f"#### {metric} Comparison")
-        metric_data = []
         
+        # Prepare data for the metric table
+        metric_data = []
         for class_name in model1_data['Class'].unique():
-            m1_value = model1_data[model1_data['Class'] == class_name][metric].iloc[0] * 100
-            m2_value = model2_data[model2_data['Class'] == class_name][metric].iloc[0] * 100
+            # Get values for both models
+            m1_value = model1_data[model1_data['Class'] == class_name][metric].iloc[0]
+            m2_value = model2_data[model2_data['Class'] == class_name][metric].iloc[0]
             diff = m1_value - m2_value
             
+            # Add to comparison data
             metric_data.append({
                 'Class': class_name,
                 model1: m1_value,
@@ -432,46 +450,92 @@ elif st.session_state.active_tab == "⚖️ Model Comparison":
                 'Difference': diff
             })
         
+        # Create DataFrame for the metric
         metric_df = pd.DataFrame(metric_data)
-        st.dataframe(style_dataframe(metric_df), use_container_width=True)
+        
+        # Style the table
+        def style_metric_table(df):
+            return df.style\
+                .format({
+                    model1: '{:.2f}%',
+                    model2: '{:.2f}%',
+                    'Difference': '{:+.2f}%'
+                })\
+                .background_gradient(
+                    cmap='RdYlGn',
+                    subset=['Difference'],
+                    vmin=-10,
+                    vmax=10
+                )\
+                .set_properties(**{
+                    'text-align': 'center',
+                    'padding': '10px',
+                    'border': '1px solid #dee2e6'
+                })\
+                .set_table_styles([
+                    {'selector': 'th', 'props': [
+                        ('background-color', '#4a90e2'),
+                        ('color', 'white'),
+                        ('font-weight', 'bold'),
+                        ('text-align', 'center'),
+                        ('padding', '10px')
+                    ]}
+                ])
+        
+        # Display the styled table
+        
+        st.dataframe(
+            style_dataframe(metric_df),
+            use_container_width=True,
+        )
+        # Add visual separator
         st.markdown("---")
     
-    # Visual comparison
+    # Visualizations section
     st.subheader("Visual Performance Analysis")
+    
+    # Metric selector for bar chart
     selected_metric = st.selectbox(
         "Select Metric for Comparison:",
         metrics,
-        key='visual_compare_metric'
+        key='compare_metric'
     )
     
-    # Prepare data for visualization
+    # Prepare data for bar chart
     comparison_data = pd.DataFrame()
+    
+    # Get data for both models
     for idx, (model_name, model_data) in enumerate([(model1, model1_data), (model2, model2_data)]):
+        # Filter out Overall classes and select relevant columns
         model_metrics = model_data[~model_data['Class'].str.contains('Overall', na=False)][['Class', selected_metric]]
         model_metrics = model_metrics.rename(columns={selected_metric: model_name})
         
+        # Merge with existing data or create new DataFrame
         if idx == 0:
             comparison_data = model_metrics
         else:
             comparison_data = comparison_data.merge(model_metrics, on='Class', how='outer')
     
-    # Bar chart
+    # Create bar chart
     fig_bar = go.Figure()
     
+    # Add bars for first model
     fig_bar.add_trace(go.Bar(
         name=model1,
         x=comparison_data['Class'],
-        y=comparison_data[model1] * 100,
+        y=comparison_data[model1],
         marker_color='rgb(55, 83, 109)'
     ))
     
+    # Add bars for second model
     fig_bar.add_trace(go.Bar(
         name=model2,
         x=comparison_data['Class'],
-        y=comparison_data[model2] * 100,
+        y=comparison_data[model2],
         marker_color='rgb(26, 118, 255)'
     ))
     
+    # Update bar chart layout
     fig_bar.update_layout(
         title=f"{selected_metric} Comparison by Class",
         xaxis_title="Class",
@@ -488,19 +552,23 @@ elif st.session_state.active_tab == "⚖️ Model Comparison":
         )
     )
     
+    # Display bar chart
     st.plotly_chart(fig_bar, use_container_width=True)
     
-    # Precision-Recall Analysis
+    # Create Precision-Recall scatter plot
     st.markdown("#### Precision-Recall Analysis")
     
+    # Filter data for scatter plot
     model1_scatter = model1_data[~model1_data['Class'].str.contains('Overall', na=False)]
     model2_scatter = model2_data[~model2_data['Class'].str.contains('Overall', na=False)]
     
+    # Create scatter plot
     fig_scatter = go.Figure()
     
+    # Add scatter points for first model
     fig_scatter.add_trace(go.Scatter(
-        x=model1_scatter['Precision'] * 100,
-        y=model1_scatter['Recall'] * 100,
+        x=model1_scatter['Precision']*100,
+        y=model1_scatter['Recall']*100,
         mode='markers+text',
         name=model1,
         text=model1_scatter['Class'],
@@ -508,9 +576,10 @@ elif st.session_state.active_tab == "⚖️ Model Comparison":
         marker=dict(size=10)
     ))
     
+    # Add scatter points for second model
     fig_scatter.add_trace(go.Scatter(
-        x=model2_scatter['Precision'] * 100,
-        y=model2_scatter['Recall'] * 100,
+        x=model2_scatter['Precision']*100,
+        y=model2_scatter['Recall']*100,
         mode='markers+text',
         name=model2,
         text=model2_scatter['Class'],
@@ -518,6 +587,7 @@ elif st.session_state.active_tab == "⚖️ Model Comparison":
         marker=dict(size=10)
     ))
     
+    # Add reference line
     fig_scatter.add_trace(go.Scatter(
         x=[0, 100],
         y=[0, 100],
@@ -526,6 +596,7 @@ elif st.session_state.active_tab == "⚖️ Model Comparison":
         showlegend=False
     ))
     
+    # Update scatter plot layout
     fig_scatter.update_layout(
         title="Precision vs Recall Analysis by Class",
         xaxis_title="Precision (%)",
@@ -542,8 +613,10 @@ elif st.session_state.active_tab == "⚖️ Model Comparison":
         )
     )
     
+    # Display scatter plot
     st.plotly_chart(fig_scatter, use_container_width=True)
 
+
 # Footer
 st.markdown("---")
 st.markdown("Dashboard created for model evaluation and comparison")