Update app.py

app.py

@@ -2,6 +2,7 @@ import gradio as gr
 import matplotlib as mpl
 import matplotlib.pyplot as plt
 import numpy as np
+from PIL import Image
 from sklearn import datasets
 from sklearn.mixture import GaussianMixture
 from sklearn.model_selection import StratifiedKFold
@@ -98,10 +99,24 @@ def classify_iris(cov_type):
     return output_path
 
 
+def update_plot(cov_type):
+    # image_path = classify_iris(cov_type)
+    # return gr.outputs.Image(image_path)
+    image_path = classify_iris(cov_type)
+    image = Image.open(image_path)
+    image_array = np.array(image)
+    return image_array
+
+
 iface = gr.Interface(
-    fn=classify_iris,
-    inputs=gr.inputs.Radio(["spherical", "diag", "tied", "full"], label="Covariance Type"),
-    outputs="image",
+    fn=update_plot,
+    inputs=gr.inputs.Radio(
+        ["spherical", "diag", "tied", "full"],
+        label="Covariance Type",
+        default="spherical"
+    ),
+    outputs=gr.outputs.Image(type="numpy"),
+    live=True,
     title="Gaussian Mixture Model Covariance",
     description="Explore different covariance types for Gaussian mixture models (GMMs) in this demonstration. GMMs are commonly used for clustering, but in this example, we compare the obtained clusters with the actual classes from the dataset. By initializing the means of the Gaussians with the means of the classes in the training set, we ensure a valid comparison. The plots show the predicted labels on both training and test data using GMMs with spherical, diagonal, full, and tied covariance matrices. Interestingly, while full covariance is expected to perform best, it may overfit small datasets and struggle to generalize to held out test data. See the original scikit-learn example for more information: https://scikit-learn.org/stable/auto_examples/mixture/plot_gmm_covariances.html",
     examples=[
@@ -112,4 +127,4 @@ iface = gr.Interface(
     ],
 )
 
-iface.launch()
+iface.launch()