mattmdjaga
commited on
Commit
•
390940a
1
Parent(s):
7c918f9
Changed checking of cached data to take into account people using the app at the same time
Browse files
app.py
CHANGED
@@ -8,12 +8,13 @@ import cv2
|
|
8 |
from typing import List
|
9 |
|
10 |
device = "cuda" if torch.cuda.is_available() else "cpu"
|
|
|
11 |
|
12 |
# Load model and processor
|
13 |
model = SamModel.from_pretrained("facebook/sam-vit-base").to(device)
|
14 |
processor = SamProcessor.from_pretrained("facebook/sam-vit-base")
|
15 |
|
16 |
-
|
17 |
|
18 |
def mask_2_dots(mask: np.ndarray) -> List[List[int]]:
|
19 |
gray = cv2.cvtColor(mask, cv2.COLOR_RGB2GRAY)
|
@@ -31,14 +32,16 @@ def mask_2_dots(mask: np.ndarray) -> List[List[int]]:
|
|
31 |
|
32 |
@torch.no_grad()
|
33 |
def foward_pass(image_input: np.ndarray, points: List[List[int]]) -> np.ndarray:
|
34 |
-
global
|
35 |
image_input = Image.fromarray(image_input)
|
36 |
inputs = processor(image_input, input_points=points, return_tensors="pt").to(device)
|
37 |
-
if not
|
38 |
embedding = model.get_image_embeddings(inputs["pixel_values"])
|
|
|
|
|
39 |
del inputs["pixel_values"]
|
40 |
|
41 |
-
outputs = model.forward(image_embeddings=
|
42 |
masks = processor.image_processor.post_process_masks(
|
43 |
outputs.pred_masks.cpu(), inputs["original_sizes"].cpu(), inputs["reshaped_input_sizes"].cpu()
|
44 |
)
|
@@ -63,9 +66,9 @@ def main_func(inputs) -> List[Image.Image]:
|
|
63 |
|
64 |
return pred_masks
|
65 |
|
66 |
-
def
|
67 |
-
global
|
68 |
-
|
69 |
|
70 |
with gr.Blocks() as demo:
|
71 |
gr.Markdown("# How to use")
|
@@ -81,6 +84,6 @@ with gr.Blocks() as demo:
|
|
81 |
image_button = gr.Button("Segment Image")
|
82 |
|
83 |
image_button.click(main_func, inputs=image_input, outputs=image_output)
|
84 |
-
image_input.upload(
|
85 |
|
86 |
demo.launch()
|
|
|
8 |
from typing import List
|
9 |
|
10 |
device = "cuda" if torch.cuda.is_available() else "cpu"
|
11 |
+
device = 'cpu'
|
12 |
|
13 |
# Load model and processor
|
14 |
model = SamModel.from_pretrained("facebook/sam-vit-base").to(device)
|
15 |
processor = SamProcessor.from_pretrained("facebook/sam-vit-base")
|
16 |
|
17 |
+
cache_data = None
|
18 |
|
19 |
def mask_2_dots(mask: np.ndarray) -> List[List[int]]:
|
20 |
gray = cv2.cvtColor(mask, cv2.COLOR_RGB2GRAY)
|
|
|
32 |
|
33 |
@torch.no_grad()
|
34 |
def foward_pass(image_input: np.ndarray, points: List[List[int]]) -> np.ndarray:
|
35 |
+
global cache_data
|
36 |
image_input = Image.fromarray(image_input)
|
37 |
inputs = processor(image_input, input_points=points, return_tensors="pt").to(device)
|
38 |
+
if not cache_data or not torch.equal(inputs['pixel_values'],cache_data[0]):
|
39 |
embedding = model.get_image_embeddings(inputs["pixel_values"])
|
40 |
+
pixels = inputs["pixel_values"]
|
41 |
+
cache_data = [pixels, embedding]
|
42 |
del inputs["pixel_values"]
|
43 |
|
44 |
+
outputs = model.forward(image_embeddings=cache_data[1], **inputs)
|
45 |
masks = processor.image_processor.post_process_masks(
|
46 |
outputs.pred_masks.cpu(), inputs["original_sizes"].cpu(), inputs["reshaped_input_sizes"].cpu()
|
47 |
)
|
|
|
66 |
|
67 |
return pred_masks
|
68 |
|
69 |
+
def reset_data():
|
70 |
+
global cache_data
|
71 |
+
cache_data = None
|
72 |
|
73 |
with gr.Blocks() as demo:
|
74 |
gr.Markdown("# How to use")
|
|
|
84 |
image_button = gr.Button("Segment Image")
|
85 |
|
86 |
image_button.click(main_func, inputs=image_input, outputs=image_output)
|
87 |
+
image_input.upload(reset_data)
|
88 |
|
89 |
demo.launch()
|