Spaces:
Running
Running
shawnljw
commited on
Commit
·
52983bc
1
Parent(s):
05ccb0c
create web ui
Browse files
app.py
ADDED
@@ -0,0 +1,98 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import gradio as gr
|
2 |
+
import numpy as np
|
3 |
+
import cv2
|
4 |
+
from tqdm import trange
|
5 |
+
|
6 |
+
class KMeansClustering():
|
7 |
+
def __init__(self, n_clusters=8, max_iter=300):
|
8 |
+
self.n_clusters = n_clusters
|
9 |
+
self.max_iter = max_iter
|
10 |
+
|
11 |
+
def fit(self, X):
|
12 |
+
self.inertia_ = float('inf')
|
13 |
+
|
14 |
+
# random init of clusters
|
15 |
+
idx = np.random.choice(range(X.shape[0]), self.n_clusters, replace=False)
|
16 |
+
self.cluster_centers_ = X[idx]
|
17 |
+
|
18 |
+
print(f'Training for {self.max_iter} epochs')
|
19 |
+
epochs = trange(self.max_iter)
|
20 |
+
for i in epochs:
|
21 |
+
distances = X[:, np.newaxis, :] - self.cluster_centers_[np.newaxis, :, :]
|
22 |
+
distances = np.linalg.norm(distances, axis=2)
|
23 |
+
self.labels_ = np.argmin(distances, axis=1)
|
24 |
+
new_inertia = np.sum(np.min(distances, axis=1) ** 2)
|
25 |
+
|
26 |
+
epochs.set_description(f'Epoch-{i+1}, Inertia-{new_inertia}')
|
27 |
+
|
28 |
+
if new_inertia < self.inertia_:
|
29 |
+
self.inertia_ = new_inertia
|
30 |
+
else:
|
31 |
+
epochs.close()
|
32 |
+
print('Early Stopping. Inertia has converged.')
|
33 |
+
break
|
34 |
+
|
35 |
+
self.cluster_centers_ = np.empty_like(self.cluster_centers_)
|
36 |
+
for cluster in range(self.n_clusters):
|
37 |
+
in_cluster = (self.labels_ == cluster)
|
38 |
+
if np.any(in_cluster):
|
39 |
+
self.cluster_centers_[cluster] = np.mean(X[in_cluster], axis=0)
|
40 |
+
else:
|
41 |
+
# cluster is empty, pick random point as next centroid
|
42 |
+
self.cluster_centers_[cluster] = X[np.random.randint(0, X.shape[0])]
|
43 |
+
|
44 |
+
return self
|
45 |
+
|
46 |
+
def predict(self, X):
|
47 |
+
distances = X[:, np.newaxis, :] - self.cluster_centers_[np.newaxis, :, :]
|
48 |
+
distances = np.linalg.norm(distances, axis=2)
|
49 |
+
labels = np.argmin(distances, axis=1)
|
50 |
+
return labels
|
51 |
+
|
52 |
+
def fit_predict(self, X):
|
53 |
+
return self.fit(X).labels_
|
54 |
+
|
55 |
+
def segment_image(image, model: KMeansClustering):
|
56 |
+
w, b, c = image.shape
|
57 |
+
image = image.reshape(w*b, c) / 255
|
58 |
+
|
59 |
+
idx = np.random.choice(range(image.shape[0]), image.shape[0]//5, replace=False)
|
60 |
+
image_subset = image[idx]
|
61 |
+
model.fit(image_subset) # fit model on 20% sample of image
|
62 |
+
|
63 |
+
labels = model.predict(image)
|
64 |
+
return labels.reshape(w,b), model
|
65 |
+
|
66 |
+
def generate_outputs(image):
|
67 |
+
model = KMeansClustering(n_clusters=24, max_iter=10)
|
68 |
+
label_map, model = segment_image(image, model)
|
69 |
+
|
70 |
+
clustered_image = model.cluster_centers_[label_map]
|
71 |
+
clustered_image = (clustered_image * 255).astype('uint8')
|
72 |
+
clustered_image = cv2.medianBlur(clustered_image,5)
|
73 |
+
edges = 255 - cv2.Canny(clustered_image, 0, 1)
|
74 |
+
edges = cv2.cvtColor(edges, cv2.COLOR_GRAY2RGB)
|
75 |
+
|
76 |
+
return [(edges, 'Coloring Page'), (clustered_image, 'Filled Picture')]
|
77 |
+
|
78 |
+
with gr.Blocks() as demo:
|
79 |
+
gr.Markdown(
|
80 |
+
"""
|
81 |
+
# image2coloringbook
|
82 |
+
|
83 |
+
(image2coloringbook)[https://github.com/ShawnLJW/image2coloringbook] is a simple tool that converts an image into a coloring book.
|
84 |
+
""")
|
85 |
+
with gr.Row():
|
86 |
+
with gr.Column():
|
87 |
+
image = gr.Image()
|
88 |
+
submit = gr.Button('Generate')
|
89 |
+
with gr.Column():
|
90 |
+
output = gr.Gallery()
|
91 |
+
submit.click(
|
92 |
+
generate_outputs,
|
93 |
+
inputs=[image],
|
94 |
+
outputs=[output]
|
95 |
+
)
|
96 |
+
|
97 |
+
if __name__ == '__main__':
|
98 |
+
demo.launch()
|