Create app.py

app.py

@@ -0,0 +1,151 @@
+import streamlit as st
+import pandas as pd
+import joblib
+import numpy as np
+import plotly.graph_objects as go
+import plotly.express as px
+
+# Set the title of the app
+st.title("Robotic Grasp Robustness Predictor")
+st.markdown("""
+This application predicts the stability/robustness of a robotic grasp based on sensor data.
+You can either upload a CSV file with sensor features or input values manually.
+""")
+
+# Load the saved model (make sure the model file 'model.pkl' is in the same directory)
+@st.cache_resource
+def load_model():
+    model = joblib.load("model.pkl")
+    return model
+
+model = load_model()
+
+# Sidebar for input selection
+st.sidebar.header("Input Options")
+input_method = st.sidebar.radio("Choose input method:", ("CSV Upload", "Manual Input"))
+
+# Define the list of features expected by the model.
+# Replace these with your actual feature names after cleaning.
+FEATURES = [
+    "H1_F1J2_pos", "H1_F1J2_vel", "H1_F1J2_eff",
+    "H1_F1J3_pos", "H1_F1J3_vel", "H1_F1J3_eff",
+    "H1_F1J1_pos", "H1_F1J1_vel", "H1_F1J1_eff",
+    "H1_F3J1_pos", "H1_F3J1_vel", "H1_F3J1_eff",
+    "H1_F3J2_pos", "H1_F3J2_vel", "H1_F3J2_eff",
+    "H1_F3J3_pos", "H1_F3J3_vel", "H1_F3J3_eff",
+    "H1_F2J1_pos", "H1_F2J1_vel", "H1_F2J1_eff",
+    "H1_F2J3_pos", "H1_F2J3_vel", "H1_F2J3_eff",
+    "H1_F2J2_pos", "H1_F2J2_vel", "H1_F2J2_eff"
+]
+
+# Option 1: CSV Upload
+if input_method == "CSV Upload":
+    st.header("Upload CSV File")
+    uploaded_file = st.file_uploader("Upload your CSV file", type=["csv"])
+    if uploaded_file is not None:
+        try:
+            input_df = pd.read_csv(uploaded_file)
+            st.success("CSV file successfully loaded!")
+            # Clean column names
+            input_df.columns = input_df.columns.str.strip()
+            missing_features = [col for col in FEATURES if col not in input_df.columns]
+            if missing_features:
+                st.error(f"The following required feature(s) are missing from the file: {', '.join(missing_features)}")
+            else:
+                # Make predictions
+                predictions = model.predict(input_df[FEATURES])
+                input_df["Predicted Robustness"] = predictions
+                
+                # Interactive histogram of predictions
+                fig_hist = px.histogram(input_df, x="Predicted Robustness", nbins=30, 
+                                        title="Distribution of Predicted Robustness")
+                st.plotly_chart(fig_hist, use_container_width=True)
+                
+                st.subheader("Predictions")
+                st.dataframe(input_df)
+        except Exception as e:
+            st.error(f"Error processing file: {e}")
+# Option 2: Manual Input
+else:
+    st.header("Manual Input")
+    st.markdown("Enter the sensor values for prediction:")
+    
+    # Group sensor inputs into sections to improve readability
+    input_data = {}
+    
+    with st.expander("F1 Joint Sensors"):
+        st.write("Sensors for Finger 1:")
+        input_data["H1_F1J2_pos"] = st.number_input("H1_F1J2_pos", value=0.0, format="%.5f")
+        input_data["H1_F1J2_vel"] = st.number_input("H1_F1J2_vel", value=0.0, format="%.5f")
+        input_data["H1_F1J2_eff"] = st.number_input("H1_F1J2_eff", value=0.0, format="%.5f")
+        
+        st.write("Sensors for Finger 3:")
+        input_data["H1_F1J3_pos"] = st.number_input("H1_F1J3_pos", value=0.0, format="%.5f")
+        input_data["H1_F1J3_vel"] = st.number_input("H1_F1J3_vel", value=0.0, format="%.5f")
+        input_data["H1_F1J3_eff"] = st.number_input("H1_F1J3_eff", value=0.0, format="%.5f")
+        
+        st.write("Sensors for Finger 1 (alternate):")
+        input_data["H1_F1J1_pos"] = st.number_input("H1_F1J1_pos", value=0.0, format="%.5f")
+        input_data["H1_F1J1_vel"] = st.number_input("H1_F1J1_vel", value=0.0, format="%.5f")
+        input_data["H1_F1J1_eff"] = st.number_input("H1_F1J1_eff", value=0.0, format="%.5f")
+        
+    with st.expander("F3 Joint Sensors"):
+        input_data["H1_F3J1_pos"] = st.number_input("H1_F3J1_pos", value=0.0, format="%.5f")
+        input_data["H1_F3J1_vel"] = st.number_input("H1_F3J1_vel", value=0.0, format="%.5f")
+        input_data["H1_F3J1_eff"] = st.number_input("H1_F3J1_eff", value=0.0, format="%.5f")
+        
+        input_data["H1_F3J2_pos"] = st.number_input("H1_F3J2_pos", value=0.0, format="%.5f")
+        input_data["H1_F3J2_vel"] = st.number_input("H1_F3J2_vel", value=0.0, format="%.5f")
+        input_data["H1_F3J2_eff"] = st.number_input("H1_F3J2_eff", value=0.0, format="%.5f")
+        
+        input_data["H1_F3J3_pos"] = st.number_input("H1_F3J3_pos", value=0.0, format="%.5f")
+        input_data["H1_F3J3_vel"] = st.number_input("H1_F3J3_vel", value=0.0, format="%.5f")
+        input_data["H1_F3J3_eff"] = st.number_input("H1_F3J3_eff", value=0.0, format="%.5f")
+        
+    with st.expander("F2 Joint Sensors"):
+        input_data["H1_F2J1_pos"] = st.number_input("H1_F2J1_pos", value=0.0, format="%.5f")
+        input_data["H1_F2J1_vel"] = st.number_input("H1_F2J1_vel", value=0.0, format="%.5f")
+        input_data["H1_F2J1_eff"] = st.number_input("H1_F2J1_eff", value=0.0, format="%.5f")
+        
+        input_data["H1_F2J3_pos"] = st.number_input("H1_F2J3_pos", value=0.0, format="%.5f")
+        input_data["H1_F2J3_vel"] = st.number_input("H1_F2J3_vel", value=0.0, format="%.5f")
+        input_data["H1_F2J3_eff"] = st.number_input("H1_F2J3_eff", value=0.0, format="%.5f")
+        
+        input_data["H1_F2J2_pos"] = st.number_input("H1_F2J2_pos", value=0.0, format="%.5f")
+        input_data["H1_F2J2_vel"] = st.number_input("H1_F2J2_vel", value=0.0, format="%.5f")
+        input_data["H1_F2J2_eff"] = st.number_input("H1_F2J2_eff", value=0.0, format="%.5f")
+    
+    # Real-time prediction as values change.
+    if st.button("Predict"):
+        input_df = pd.DataFrame([input_data])
+        prediction = model.predict(input_df)
+        st.success(f"The predicted grasp robustness is: {prediction[0]:.3f}")
+        
+        # Create a gauge chart (using Plotly) to visualize the prediction
+        # (Customize the min/max values based on your domain; below, 0-100 is used as an example.)
+        fig_gauge = go.Figure(go.Indicator(
+            mode = "gauge+number",
+            value = prediction[0],
+            title = {'text': "Grasp Robustness"},
+            gauge = {'axis': {'range': [None, 100]},
+                     'bar': {'color': "darkblue"},
+                     'steps' : [
+                         {'range': [0, 30], 'color': "red"},
+                         {'range': [30, 70], 'color': "yellow"},
+                         {'range': [70, 100], 'color': "green"}],
+                     }
+        ))
+        st.plotly_chart(fig_gauge, use_container_width=True)
+        
+        # If the model supports feature importances (e.g. RandomForest), display them
+        try:
+            feature_importances = model.named_steps['model'].feature_importances_
+            imp_df = pd.DataFrame({
+                'Feature': FEATURES,
+                'Importance': feature_importances
+            }).sort_values(by='Importance', ascending=False)
+            st.subheader("Feature Importances")
+            fig_imp = px.bar(imp_df, x='Feature', y='Importance', title="Feature Importances")
+            st.plotly_chart(fig_imp, use_container_width=True)
+        except Exception as ex:
+            st.info("Feature importance is not available for this model.")