Spaces:
Runtime error
Runtime error
Update app.py
Browse files
app.py
CHANGED
@@ -6,14 +6,12 @@ from PIL import Image, ImageDraw
|
|
6 |
import numpy as np
|
7 |
from transformers import AutoImageProcessor, Mask2FormerForUniversalSegmentation
|
8 |
|
9 |
-
import numpy as np
|
10 |
from scipy.ndimage import center_of_mass
|
11 |
|
12 |
-
|
13 |
def combine_ims(im1, im2, val=128):
|
14 |
-
|
15 |
-
|
16 |
-
|
17 |
|
18 |
def get_class_centers(segmentation_mask, class_dict):
|
19 |
segmentation_mask = segmentation_mask.numpy() + 1
|
@@ -24,44 +22,47 @@ def get_class_centers(segmentation_mask, class_dict):
|
|
24 |
|
25 |
class_centers[class_index] = center_of_mass_list
|
26 |
|
27 |
-
class_centers = {k:list(map(int, v)) for k,v in class_centers.items() if not np.isnan(sum(v))}
|
28 |
return class_centers
|
29 |
|
30 |
def visualize_mask(predicted_semantic_map, class_ids, class_colors):
|
31 |
-
|
32 |
-
|
33 |
-
|
34 |
-
|
35 |
-
|
36 |
-
colors = class_colors[class_ids[color_indexes]]
|
37 |
-
output = colors.reshape(h, w, 3).astype(np.uint8)
|
38 |
-
image_mask = Image.fromarray(output)
|
39 |
-
return image_mask
|
40 |
|
|
|
|
|
|
|
|
|
41 |
|
42 |
def get_out_image(image, predicted_semantic_map):
|
43 |
-
|
44 |
-
|
45 |
-
|
46 |
-
|
47 |
-
|
48 |
-
|
|
|
|
|
|
|
|
|
49 |
|
50 |
-
|
51 |
|
52 |
def gradio_process(image):
|
53 |
-
|
54 |
|
55 |
-
|
56 |
-
|
57 |
|
58 |
-
|
59 |
|
60 |
-
|
61 |
-
|
62 |
|
63 |
with open('ade20k_classes.pickle', 'rb') as f:
|
64 |
-
|
65 |
class_names, class_ids, class_colors = np.array(class_names), np.array(class_ids), np.array(class_colors)
|
66 |
class_dict = dict(zip(class_ids, class_names))
|
67 |
|
@@ -73,11 +74,10 @@ model.eval()
|
|
73 |
demo = gr.Interface(
|
74 |
gradio_process,
|
75 |
inputs=gr.inputs.Image(type="pil"),
|
76 |
-
outputs=gr.outputs.Image(type="pil"),
|
77 |
title="Semantic Segmentation",
|
78 |
examples=glob.glob('./examples/*.jpg'),
|
79 |
allow_flagging="never",
|
80 |
-
|
81 |
)
|
82 |
|
83 |
demo.launch()
|
|
|
6 |
import numpy as np
|
7 |
from transformers import AutoImageProcessor, Mask2FormerForUniversalSegmentation
|
8 |
|
|
|
9 |
from scipy.ndimage import center_of_mass
|
10 |
|
|
|
11 |
def combine_ims(im1, im2, val=128):
|
12 |
+
p = Image.new("L", im1.size, val)
|
13 |
+
im = Image.composite(im1, im2, p)
|
14 |
+
return im
|
15 |
|
16 |
def get_class_centers(segmentation_mask, class_dict):
|
17 |
segmentation_mask = segmentation_mask.numpy() + 1
|
|
|
22 |
|
23 |
class_centers[class_index] = center_of_mass_list
|
24 |
|
25 |
+
class_centers = {k: list(map(int, v)) for k, v in class_centers.items() if not np.isnan(sum(v))}
|
26 |
return class_centers
|
27 |
|
28 |
def visualize_mask(predicted_semantic_map, class_ids, class_colors):
|
29 |
+
h, w = predicted_semantic_map.shape
|
30 |
+
color_indexes = np.zeros((h, w), dtype=np.uint8)
|
31 |
+
color_indexes[:] = predicted_semantic_map.numpy()
|
32 |
+
color_indexes = color_indexes.flatten()
|
|
|
|
|
|
|
|
|
|
|
33 |
|
34 |
+
colors = class_colors[class_ids[color_indexes]]
|
35 |
+
output = colors.reshape(h, w, 3).astype(np.uint8)
|
36 |
+
image_mask = Image.fromarray(output)
|
37 |
+
return image_mask
|
38 |
|
39 |
def get_out_image(image, predicted_semantic_map):
|
40 |
+
class_centers = get_class_centers(predicted_semantic_map, class_dict)
|
41 |
+
mask = visualize_mask(predicted_semantic_map, class_ids, class_colors)
|
42 |
+
image_mask = combine_ims(image, mask, val=128)
|
43 |
+
draw = ImageDraw.Draw(image_mask)
|
44 |
+
|
45 |
+
extracted_tags = []
|
46 |
+
for id, (y, x) in class_centers.items():
|
47 |
+
class_name = str(class_names[id - 1])
|
48 |
+
extracted_tags.append({"class_name": class_name, "coordinates": (x, y)})
|
49 |
+
draw.text((x, y), class_name, fill='black')
|
50 |
|
51 |
+
return image_mask, extracted_tags
|
52 |
|
53 |
def gradio_process(image):
|
54 |
+
inputs = processor(images=image, return_tensors="pt")
|
55 |
|
56 |
+
with torch.no_grad():
|
57 |
+
outputs = model(**inputs)
|
58 |
|
59 |
+
predicted_semantic_map = processor.post_process_semantic_segmentation(outputs, target_sizes=[image.size[::-1]])[0]
|
60 |
|
61 |
+
out_image, extracted_tags = get_out_image(image, predicted_semantic_map)
|
62 |
+
return out_image, extracted_tags
|
63 |
|
64 |
with open('ade20k_classes.pickle', 'rb') as f:
|
65 |
+
class_names, class_ids, class_colors = pickle.load(f)
|
66 |
class_names, class_ids, class_colors = np.array(class_names), np.array(class_ids), np.array(class_colors)
|
67 |
class_dict = dict(zip(class_ids, class_names))
|
68 |
|
|
|
74 |
demo = gr.Interface(
|
75 |
gradio_process,
|
76 |
inputs=gr.inputs.Image(type="pil"),
|
77 |
+
outputs=[gr.outputs.Image(type="pil"), gr.outputs.JSON()],
|
78 |
title="Semantic Segmentation",
|
79 |
examples=glob.glob('./examples/*.jpg'),
|
80 |
allow_flagging="never",
|
|
|
81 |
)
|
82 |
|
83 |
demo.launch()
|