Update to latest Gradio and fix deprecated APIs

- Update sdk_version from 2.8.12 to 5.29.0
- Replace deprecated gr.inputs/gr.outputs with modern Gradio components
- Replace removed huggingface_hub.keras_mixin.from_pretrained_keras with snapshot_download + tf.keras.models.load_model
- Replace tensorflow.keras.preprocessing.image with numpy operations
- Fix cv2.merge to use 2D arrays instead of 3D
- Cast slider values to int where needed
- Add huggingface_hub to requirements.txt

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

README.md

@@ -4,7 +4,7 @@ emoji: 👀
 colorFrom: blue
 colorTo: gray
 sdk: gradio
-sdk_version: 2.8.12
+sdk_version: 5.29.0
 app_file: app.py
 pinned: false
 ---

app.py

@@ -4,15 +4,10 @@ from sklearn import datasets
 import numpy as np
 from matplotlib import pyplot as plt
 from scipy import ndimage
-from skimage import measure, color, io
-from tensorflow.keras.preprocessing import image
-from scipy import ndimage
-import skimage.io as io
-import skimage.transform as trans
-import numpy as np
+from skimage import measure, color
 import tensorflow as tf
 import gradio as gr
-from huggingface_hub.keras_mixin import from_pretrained_keras
+from huggingface_hub import snapshot_download
 from itertools import cycle, islice
 
 
@@ -49,7 +44,8 @@ def create_input_image(data, visualize=False):
 
   return input
 
-model = from_pretrained_keras("tareknaous/unet-visual-clustering")
+model_path = snapshot_download("tareknaous/unet-visual-clustering")
+model = tf.keras.models.load_model(model_path)
 
 
 def get_instances(prediction, data, max_filter_size=1):
@@ -60,7 +56,9 @@ def get_instances(prediction, data, max_filter_size=1):
   prediction[prediction == 4] = 255
 
   #Convert to 8-bit image
-  prediction = image.img_to_array(prediction, dtype='uint8')
+  prediction = np.uint8(prediction)
+  if prediction.ndim == 2:
+    prediction = prediction[..., np.newaxis]
   
   #Get 1 color channel
   cells=prediction[:,:,0]
@@ -83,7 +81,7 @@ def get_instances(prediction, data, max_filter_size=1):
   markers[unknown==255] = 0
 
   #Watershed
-  img = cv2.merge((prediction,prediction,prediction))
+  img = cv2.merge((cells, cells, cells))
   markers = cv2.watershed(img,markers)
   img[markers == -1] = [0,255,255]  
 
@@ -125,6 +123,11 @@ def get_instances(prediction, data, max_filter_size=1):
 
 
 def visual_clustering(cluster_type, num_clusters, num_samples, noise, random_state, median_kernel_size, max_kernel_size):
+  num_clusters = int(num_clusters)
+  num_samples = int(num_samples)
+  random_state = int(random_state)
+  median_kernel_size = int(median_kernel_size)
+  max_kernel_size = int(max_kernel_size)
 
   NUM_CLUSTERS = num_clusters
   CLUSTER_STD = 4 * np.ones(NUM_CLUSTERS)
@@ -197,25 +200,21 @@ Gradio Demo for Visual Clustering on synthetic datasets.
 '''
 
 iface = gr.Interface(
-    
-  fn=visual_clustering, 
-
+  fn=visual_clustering,
   inputs=[
-          gr.inputs.Dropdown(["blobs", "varied blobs",  "aniso", "noisy moons", "noisy circles" ]),
-          gr.inputs.Slider(1, 10, step=1, label='Number of Clusters'),
-          gr.inputs.Slider(10000, 1000000, step=10000, label='Number of Samples'),
-          gr.inputs.Slider(0.03, 0.1, step=0.01, label='Noise'),
-          gr.inputs.Slider(1, 100, step=1, label='Random State'),
-          gr.inputs.Slider(1, 100, step=1, label='Denoising Filter Kernel Size'),
-          gr.inputs.Slider(1,100, step=1, label='Max Filter Kernel Size')
-          ],
-
+    gr.Dropdown(choices=["blobs", "varied blobs", "aniso", "noisy moons", "noisy circles"], label="Cluster Type"),
+    gr.Slider(minimum=1, maximum=10, step=1, label="Number of Clusters"),
+    gr.Slider(minimum=10000, maximum=1000000, step=10000, label="Number of Samples"),
+    gr.Slider(minimum=0.03, maximum=0.1, step=0.01, label="Noise"),
+    gr.Slider(minimum=1, maximum=100, step=1, label="Random State"),
+    gr.Slider(minimum=1, maximum=100, step=1, label="Denoising Filter Kernel Size"),
+    gr.Slider(minimum=1, maximum=100, step=1, label="Max Filter Kernel Size"),
+  ],
   outputs=[
-           gr.outputs.Image(type='plot', label='Dataset'),
-           gr.outputs.Image(type='plot', label='Clustering Result')
-           ],
-           
+    gr.Plot(label="Dataset"),
+    gr.Plot(label="Clustering Result"),
+  ],
   title=title,
   description=description,
-           )
-iface.launch(debug=True)
+)
+iface.launch()

requirements.txt

@@ -6,4 +6,5 @@ scipy
 tensorflow
 matplotlib
 numpy
-opencv-python
+opencv-python
+huggingface_hub