Update app.py

app.py

@@ -3,6 +3,7 @@ import numpy as np
 import matplotlib.pyplot as plt
 from sklearn.ensemble import RandomForestRegressor, RandomForestClassifier
 from sklearn.metrics import mean_squared_error, accuracy_score
+import imageio, os, tempfile
 
 # ------------------------------------------------
 # Helpers
@@ -33,17 +34,13 @@ def generate_classification(n_points, noise):
 
 
 # ------------------------------------------------
-# 2D Random Forest Regression + Trees + Importance
+# 2D Random Forest Regression
 # ------------------------------------------------
 
 def rf_2d(n_points, noise, n_estimators, max_depth, show_trees):
     X, y = generate_2d_regression(n_points, noise)
 
-    model = RandomForestRegressor(
-        n_estimators=n_estimators,
-        max_depth=max_depth,
-        random_state=42
-    )
+    model = RandomForestRegressor(n_estimators=n_estimators, max_depth=max_depth, random_state=42)
     model.fit(X, y)
     y_pred = model.predict(X)
 
@@ -53,7 +50,6 @@ def rf_2d(n_points, noise, n_estimators, max_depth, show_trees):
     ax.scatter(X, y, s=20, label="Data")
     ax.plot(X, y_pred, linewidth=2, label="RF Avg")
 
-    # Individual trees
     if show_trees:
         for tree in model.estimators_[:5]:
             ax.plot(X, tree.predict(X), linewidth=1, alpha=0.5)
@@ -61,7 +57,6 @@ def rf_2d(n_points, noise, n_estimators, max_depth, show_trees):
     ax.set_title("2D Random Forest Regression")
     ax.legend()
 
-    # Feature importance
     imp_fig, imp_ax = plt.subplots(figsize=(4, 3))
     imp_ax.bar(["x"], model.feature_importances_)
     imp_ax.set_title("Feature Importance")
@@ -70,64 +65,67 @@ def rf_2d(n_points, noise, n_estimators, max_depth, show_trees):
 
 
 # ------------------------------------------------
-# 3D Random Forest Regression + Rotation
+# 3D Random Forest Regression with GIF rotation
 # ------------------------------------------------
 
 def rf_3d(n_points, noise, n_estimators, max_depth, rotate):
     X1, X2, X_flat, Z, Z_flat = generate_3d_regression(n_points, noise)
 
-    model = RandomForestRegressor(
-        n_estimators=n_estimators,
-        max_depth=max_depth,
-        random_state=42
-    )
+    model = RandomForestRegressor(n_estimators=n_estimators, max_depth=max_depth, random_state=42)
     model.fit(X_flat, Z_flat)
 
     Z_pred = model.predict(X_flat).reshape(X1.shape)
     mse = mean_squared_error(Z_flat, Z_pred.ravel())
 
-    frames = []
+    # Static plot
+    fig = plt.figure(figsize=(5, 4))
+    ax = fig.add_subplot(111, projection="3d")
+    idx = np.random.choice(len(Z_flat), min(400, len(Z_flat)), replace=False)
+    ax.scatter(X_flat[idx, 0], X_flat[idx, 1], Z_flat[idx], s=8, alpha=0.3)
+    ax.plot_surface(X1, X2, Z_pred, alpha=0.7)
+    ax.set_title("3D Random Forest Surface")
 
-    angles = range(0, 360, 10) if rotate else [45]
+    gif_path = None
 
-    for angle in angles:
-        fig = plt.figure(figsize=(5, 4))
-        ax = fig.add_subplot(111, projection="3d")
+    # Create rotating GIF if enabled
+    if rotate:
+        tmpdir = tempfile.mkdtemp()
+        images = []
+        for angle in range(0, 360, 10):
+            fig_rot = plt.figure(figsize=(5, 4))
+            ax_rot = fig_rot.add_subplot(111, projection="3d")
+            ax_rot.scatter(X_flat[idx, 0], X_flat[idx, 1], Z_flat[idx], s=8, alpha=0.3)
+            ax_rot.plot_surface(X1, X2, Z_pred, alpha=0.7)
+            ax_rot.view_init(elev=25, azim=angle)
 
-        idx = np.random.choice(len(Z_flat), min(400, len(Z_flat)), replace=False)
-        ax.scatter(X_flat[idx, 0], X_flat[idx, 1], Z_flat[idx], s=8, alpha=0.3)
+            frame_path = os.path.join(tmpdir, f"frame_{angle}.png")
+            fig_rot.savefig(frame_path)
+            plt.close(fig_rot)
+            images.append(imageio.imread(frame_path))
 
-        ax.plot_surface(X1, X2, Z_pred, alpha=0.7)
-        ax.view_init(elev=25, azim=angle)
-        ax.set_title("3D Random Forest Surface")
+        gif_path = os.path.join(tmpdir, "rotation.gif")
+        imageio.mimsave(gif_path, images, duration=0.08)
 
-        frames.append(fig)
-
-    return frames, f"MSE: {mse:.4f}"
+    return fig, gif_path, f"MSE: {mse:.4f}"
 
 
 # ------------------------------------------------
-# Classification Version
+# Classification
 # ------------------------------------------------
 
 def rf_classification(n_points, noise, n_estimators, max_depth):
     X, y = generate_classification(n_points, noise)
 
-    model = RandomForestClassifier(
-        n_estimators=n_estimators,
-        max_depth=max_depth,
-        random_state=42
-    )
+    model = RandomForestClassifier(n_estimators=n_estimators, max_depth=max_depth, random_state=42)
     model.fit(X, y)
 
     y_pred = model.predict(X)
     acc = accuracy_score(y, y_pred)
 
     fig, ax = plt.subplots(figsize=(5, 4))
-    scatter = ax.scatter(X[:, 0], X[:, 1], c=y_pred, s=20)
+    ax.scatter(X[:, 0], X[:, 1], c=y_pred, s=20)
     ax.set_title("Random Forest Classification")
 
-    # Feature importance
     imp_fig, imp_ax = plt.subplots(figsize=(4, 3))
     imp_ax.bar(["x1", "x2"], model.feature_importances_)
     imp_ax.set_title("Feature Importance")
@@ -136,17 +134,13 @@ def rf_classification(n_points, noise, n_estimators, max_depth):
 
 
 # ------------------------------------------------
-# Gradio UI (Auto‑run on input change)
+# Gradio UI
 # ------------------------------------------------
 
 with gr.Blocks() as demo:
-    gr.Markdown("# 🌲 Random Forest Visualizer (Full Interactive)")
+    gr.Markdown("# 🌲 Random Forest Visualizer (Fixed + Rotation)")
 
-    mode = gr.Radio(
-        ["2D Regression", "3D Regression", "Classification"],
-        value="2D Regression",
-        label="Mode"
-    )
+    mode = gr.Radio(["2D Regression", "3D Regression", "Classification"], value="2D Regression", label="Mode")
 
     n_points = gr.Slider(50, 200, value=100, step=10, label="Data Points")
     noise = gr.Slider(0.0, 5.0, value=1.0, label="Noise")
@@ -154,29 +148,32 @@ with gr.Blocks() as demo:
     max_depth = gr.Slider(2, 20, value=8, step=1, label="Max Depth")
 
     show_trees = gr.Checkbox(label="Show Individual Trees (2D)", value=False)
-    rotate = gr.Checkbox(label="Rotate 3D", value=False)
+    rotate = gr.Checkbox(label="Rotate 3D (GIF)", value=False)
 
     plot = gr.Plot()
     plot2 = gr.Plot()
+    gif = gr.Image(label="3D Rotation")
     metric = gr.Markdown()
 
     def run(mode, n_points, noise, n_estimators, max_depth, show_trees, rotate):
         if mode == "2D Regression":
             fig1, fig2, text = rf_2d(n_points, noise, n_estimators, max_depth, show_trees)
-            return fig1, fig2, text
+            return fig1, fig2, None, text
+
         elif mode == "3D Regression":
-            frames, text = rf_3d(n_points, noise, n_estimators, max_depth, rotate)
-            return frames[0], None, text
+            fig, gif_path, text = rf_3d(n_points, noise, n_estimators, max_depth, rotate)
+            return fig, None, gif_path, text
+
         else:
             fig1, fig2, text = rf_classification(n_points, noise, n_estimators, max_depth)
-            return fig1, fig2, text
+            return fig1, fig2, None, text
 
     inputs = [mode, n_points, noise, n_estimators, max_depth, show_trees, rotate]
 
     for inp in inputs:
-        inp.change(run, inputs=inputs, outputs=[plot, plot2, metric])
+        inp.change(run, inputs=inputs, outputs=[plot, plot2, gif, metric])
 
-    demo.load(run, inputs=inputs, outputs=[plot, plot2, metric])
+    demo.load(run, inputs=inputs, outputs=[plot, plot2, gif, metric])
 
 
 demo.launch()