Spaces:
Running
Running
Martin Tomov
commited on
Commit
β’
9fcd716
1
Parent(s):
81b2e04
Update gsl_utils.py
Browse files- gsl_utils.py +15 -14
gsl_utils.py
CHANGED
@@ -1,4 +1,5 @@
|
|
1 |
# GSL
|
|
|
2 |
import os
|
3 |
import torch
|
4 |
import numpy as np
|
@@ -15,12 +16,12 @@ def load_groundingdino_model(device='cpu'):
|
|
15 |
model = pipeline(model="IDEA-Research/grounding-dino-base", task="zero-shot-object-detection", device=device)
|
16 |
return model
|
17 |
|
18 |
-
def load_sam_model(device='cpu'):
|
19 |
-
sam_model = build_sam(checkpoint=
|
20 |
return SamPredictor(sam_model)
|
21 |
|
22 |
groundingdino_model = load_groundingdino_model(device=device)
|
23 |
-
sam_predictor = load_sam_model(device=device)
|
24 |
simple_lama = SimpleLama()
|
25 |
|
26 |
def detect(image, model, text_prompt='insect . flower . cloud', box_threshold=0.15, text_threshold=0.15):
|
@@ -63,11 +64,11 @@ def dilate_mask(mask, dilate_factor=15):
|
|
63 |
return mask
|
64 |
|
65 |
def gsl_process_image(image):
|
66 |
-
#
|
67 |
if not isinstance(image, np.ndarray):
|
68 |
image = np.array(image)
|
69 |
|
70 |
-
# load
|
71 |
image_pil = Image.fromarray(image)
|
72 |
|
73 |
detected_boxes = detect(image_pil, groundingdino_model)
|
@@ -84,16 +85,16 @@ def gsl_process_image(image):
|
|
84 |
annotated_frame_with_mask = draw_mask(final_mask, image)
|
85 |
|
86 |
mask = final_mask.numpy()
|
87 |
-
mask
|
88 |
-
mask
|
89 |
-
dilated_image_mask_pil
|
90 |
|
91 |
-
result
|
92 |
|
93 |
-
diff
|
94 |
-
threshold
|
95 |
-
diff2
|
96 |
-
img3
|
97 |
-
diff3
|
98 |
|
99 |
return diff3
|
|
|
1 |
# GSL
|
2 |
+
|
3 |
import os
|
4 |
import torch
|
5 |
import numpy as np
|
|
|
16 |
model = pipeline(model="IDEA-Research/grounding-dino-base", task="zero-shot-object-detection", device=device)
|
17 |
return model
|
18 |
|
19 |
+
def load_sam_model(checkpoint_path, device='cpu'):
|
20 |
+
sam_model = build_sam(checkpoint=checkpoint_path).to(device)
|
21 |
return SamPredictor(sam_model)
|
22 |
|
23 |
groundingdino_model = load_groundingdino_model(device=device)
|
24 |
+
sam_predictor = load_sam_model(checkpoint_path="models/sam_vit_h_4b8939.pth", device=device)
|
25 |
simple_lama = SimpleLama()
|
26 |
|
27 |
def detect(image, model, text_prompt='insect . flower . cloud', box_threshold=0.15, text_threshold=0.15):
|
|
|
64 |
return mask
|
65 |
|
66 |
def gsl_process_image(image):
|
67 |
+
# numpy array
|
68 |
if not isinstance(image, np.ndarray):
|
69 |
image = np.array(image)
|
70 |
|
71 |
+
# load image as a PIL
|
72 |
image_pil = Image.fromarray(image)
|
73 |
|
74 |
detected_boxes = detect(image_pil, groundingdino_model)
|
|
|
85 |
annotated_frame_with_mask = draw_mask(final_mask, image)
|
86 |
|
87 |
mask = final_mask.numpy()
|
88 |
+
mask is mask.astype(np.uint8) * 255
|
89 |
+
mask is dilate_mask(mask)
|
90 |
+
dilated_image_mask_pil is Image.fromarray(mask)
|
91 |
|
92 |
+
result is simple_lama(image, dilated_image_mask_pil)
|
93 |
|
94 |
+
diff is ImageChops.difference(result, Image.fromarray(image))
|
95 |
+
threshold is 7
|
96 |
+
diff2 is diff.convert('L').point(lambda p: 255 if p > threshold else 0).convert('1')
|
97 |
+
img3 is Image.new('RGB', Image.fromarray(image).size, (255, 236, 10))
|
98 |
+
diff3 is Image.composite(Image.fromarray(image), img3, diff2)
|
99 |
|
100 |
return diff3
|