Thiago Hersan
commited on
Commit
•
7bb7f6b
1
Parent(s):
15625a2
adds examples
Browse files- README.md +1 -1
- app.py +22 -7
- examples/map-000.jpg +0 -0
- examples/map-010.jpg +0 -0
- examples/map-018.jpg +0 -0
- examples/map-114.jpg +0 -0
- requirements.txt +3 -2
README.md
CHANGED
@@ -4,7 +4,7 @@ emoji: 🥦
|
|
4 |
colorFrom: pink
|
5 |
colorTo: indigo
|
6 |
sdk: gradio
|
7 |
-
sdk_version: 3.16.
|
8 |
app_file: app.py
|
9 |
models:
|
10 |
- "facebook/maskformer-swin-large-coco"
|
|
|
4 |
colorFrom: pink
|
5 |
colorTo: indigo
|
6 |
sdk: gradio
|
7 |
+
sdk_version: 3.16.2
|
8 |
app_file: app.py
|
9 |
models:
|
10 |
- "facebook/maskformer-swin-large-coco"
|
app.py
CHANGED
@@ -1,5 +1,7 @@
|
|
|
|
1 |
import gradio as gr
|
2 |
import numpy as np
|
|
|
3 |
from transformers import MaskFormerFeatureExtractor, MaskFormerForInstanceSegmentation
|
4 |
|
5 |
|
@@ -8,6 +10,8 @@ from transformers import MaskFormerFeatureExtractor, MaskFormerForInstanceSegmen
|
|
8 |
feature_extractor = MaskFormerFeatureExtractor.from_pretrained("facebook/maskformer-swin-large-coco")
|
9 |
model = MaskFormerForInstanceSegmentation.from_pretrained("facebook/maskformer-swin-large-coco")
|
10 |
|
|
|
|
|
11 |
def visualize_instance_seg_mask(img_in, mask, id2label):
|
12 |
img_out = np.zeros((mask.shape[0], mask.shape[1], 3))
|
13 |
image_total_pixels = mask.shape[0] * mask.shape[1]
|
@@ -28,7 +32,7 @@ def visualize_instance_seg_mask(img_in, mask, id2label):
|
|
28 |
img_out[i, j, :] = id2color[mask[i, j]]
|
29 |
id2count[mask[i, j]] = id2count[mask[i, j]] + 1
|
30 |
|
31 |
-
image_res = (0.5 * img_in + 0.5 * img_out)
|
32 |
|
33 |
vegetation_count = sum([id2count[id] for id in label_ids if id2label[id] in vegetation_labels])
|
34 |
|
@@ -47,28 +51,39 @@ def visualize_instance_seg_mask(img_in, mask, id2label):
|
|
47 |
f"{(100 * vegetation_count / image_total_pixels):.2f} %",
|
48 |
f"{np.sqrt(vegetation_count / image_total_pixels):.2f} m"]]
|
49 |
|
50 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
51 |
|
52 |
|
53 |
-
def query_image(
|
|
|
54 |
img_size = (img.shape[0], img.shape[1])
|
55 |
inputs = feature_extractor(images=img, return_tensors="pt")
|
56 |
outputs = model(**inputs)
|
57 |
results = feature_extractor.post_process_semantic_segmentation(outputs=outputs, target_sizes=[img_size])[0]
|
58 |
-
|
59 |
-
return
|
60 |
|
61 |
|
62 |
demo = gr.Interface(
|
63 |
query_image,
|
64 |
-
inputs=[gr.Image(label="Input Image")],
|
65 |
outputs=[
|
66 |
gr.Image(label="Vegetation"),
|
67 |
gr.DataFrame(label="Info", headers=["Object Label", "Pixel Percent", "Square Length"])
|
68 |
],
|
69 |
title="Maskformer (large-coco)",
|
70 |
allow_flagging="never",
|
71 |
-
analytics_enabled=None
|
|
|
|
|
72 |
)
|
73 |
|
74 |
demo.launch(show_api=False)
|
|
|
1 |
+
import glob
|
2 |
import gradio as gr
|
3 |
import numpy as np
|
4 |
+
from PIL import Image
|
5 |
from transformers import MaskFormerFeatureExtractor, MaskFormerForInstanceSegmentation
|
6 |
|
7 |
|
|
|
10 |
feature_extractor = MaskFormerFeatureExtractor.from_pretrained("facebook/maskformer-swin-large-coco")
|
11 |
model = MaskFormerForInstanceSegmentation.from_pretrained("facebook/maskformer-swin-large-coco")
|
12 |
|
13 |
+
example_images = sorted(glob.glob('examples/map*.jpg'))
|
14 |
+
|
15 |
def visualize_instance_seg_mask(img_in, mask, id2label):
|
16 |
img_out = np.zeros((mask.shape[0], mask.shape[1], 3))
|
17 |
image_total_pixels = mask.shape[0] * mask.shape[1]
|
|
|
32 |
img_out[i, j, :] = id2color[mask[i, j]]
|
33 |
id2count[mask[i, j]] = id2count[mask[i, j]] + 1
|
34 |
|
35 |
+
image_res = (0.5 * img_in + 0.5 * img_out).astype(np.uint8)
|
36 |
|
37 |
vegetation_count = sum([id2count[id] for id in label_ids if id2label[id] in vegetation_labels])
|
38 |
|
|
|
51 |
f"{(100 * vegetation_count / image_total_pixels):.2f} %",
|
52 |
f"{np.sqrt(vegetation_count / image_total_pixels):.2f} m"]]
|
53 |
|
54 |
+
dataframe = dataframe_vegetation_total
|
55 |
+
if len(dataframe) < 1:
|
56 |
+
dataframe = [[
|
57 |
+
f"",
|
58 |
+
f"{(0):.2f} %",
|
59 |
+
f"{(0):.2f} m"
|
60 |
+
]]
|
61 |
+
|
62 |
+
return image_res, dataframe
|
63 |
|
64 |
|
65 |
+
def query_image(image_path):
|
66 |
+
img = np.array(Image.open(image_path))
|
67 |
img_size = (img.shape[0], img.shape[1])
|
68 |
inputs = feature_extractor(images=img, return_tensors="pt")
|
69 |
outputs = model(**inputs)
|
70 |
results = feature_extractor.post_process_semantic_segmentation(outputs=outputs, target_sizes=[img_size])[0]
|
71 |
+
mask_img, dataframe = visualize_instance_seg_mask(img, results.numpy(), model.config.id2label)
|
72 |
+
return mask_img, dataframe
|
73 |
|
74 |
|
75 |
demo = gr.Interface(
|
76 |
query_image,
|
77 |
+
inputs=[gr.Image(type="filepath", label="Input Image")],
|
78 |
outputs=[
|
79 |
gr.Image(label="Vegetation"),
|
80 |
gr.DataFrame(label="Info", headers=["Object Label", "Pixel Percent", "Square Length"])
|
81 |
],
|
82 |
title="Maskformer (large-coco)",
|
83 |
allow_flagging="never",
|
84 |
+
analytics_enabled=None,
|
85 |
+
examples=example_images,
|
86 |
+
cache_examples=True
|
87 |
)
|
88 |
|
89 |
demo.launch(show_api=False)
|
examples/map-000.jpg
ADDED
![]() |
examples/map-010.jpg
ADDED
![]() |
examples/map-018.jpg
ADDED
![]() |
examples/map-114.jpg
ADDED
![]() |
requirements.txt
CHANGED
@@ -1,3 +1,4 @@
|
|
1 |
-
|
|
|
2 |
torch
|
3 |
-
|
|
|
1 |
+
Pillow
|
2 |
+
scipy
|
3 |
torch
|
4 |
+
transformers
|