mattmdjaga
commited on
Commit
•
fd219d5
1
Parent(s):
21232f6
Added no grad and storing embeddings
Browse files- app.py +13 -2
- requirements.txt +2 -1
app.py
CHANGED
@@ -13,6 +13,8 @@ device = "cuda" if torch.cuda.is_available() else "cpu"
|
|
13 |
model = SamModel.from_pretrained("facebook/sam-vit-base").to(device)
|
14 |
processor = SamProcessor.from_pretrained("facebook/sam-vit-base")
|
15 |
|
|
|
|
|
16 |
def mask_2_dots(mask: np.ndarray) -> List[List[int]]:
|
17 |
gray = cv2.cvtColor(mask, cv2.COLOR_RGB2GRAY)
|
18 |
_, thresh = cv2.threshold(gray, 127, 255, 0)
|
@@ -27,11 +29,16 @@ def mask_2_dots(mask: np.ndarray) -> List[List[int]]:
|
|
27 |
points.append([cx, cy])
|
28 |
return [points]
|
29 |
|
|
|
30 |
def foward_pass(image_input: np.ndarray, points: List[List[int]]) -> np.ndarray:
|
|
|
31 |
image_input = Image.fromarray(image_input)
|
32 |
-
|
33 |
inputs = processor(image_input, input_points=points, return_tensors="pt").to(device)
|
34 |
-
|
|
|
|
|
|
|
|
|
35 |
masks = processor.image_processor.post_process_masks(
|
36 |
outputs.pred_masks.cpu(), inputs["original_sizes"].cpu(), inputs["reshaped_input_sizes"].cpu()
|
37 |
)
|
@@ -56,6 +63,9 @@ def main_func(inputs) -> List[Image.Image]:
|
|
56 |
|
57 |
return pred_masks
|
58 |
|
|
|
|
|
|
|
59 |
|
60 |
with gr.Blocks() as demo:
|
61 |
gr.Markdown("# How to use")
|
@@ -71,5 +81,6 @@ with gr.Blocks() as demo:
|
|
71 |
image_button = gr.Button("Segment Image")
|
72 |
|
73 |
image_button.click(main_func, inputs=image_input, outputs=image_output)
|
|
|
74 |
|
75 |
demo.launch()
|
|
|
13 |
model = SamModel.from_pretrained("facebook/sam-vit-base").to(device)
|
14 |
processor = SamProcessor.from_pretrained("facebook/sam-vit-base")
|
15 |
|
16 |
+
embedding = None
|
17 |
+
|
18 |
def mask_2_dots(mask: np.ndarray) -> List[List[int]]:
|
19 |
gray = cv2.cvtColor(mask, cv2.COLOR_RGB2GRAY)
|
20 |
_, thresh = cv2.threshold(gray, 127, 255, 0)
|
|
|
29 |
points.append([cx, cy])
|
30 |
return [points]
|
31 |
|
32 |
+
@torch.no_grad()
|
33 |
def foward_pass(image_input: np.ndarray, points: List[List[int]]) -> np.ndarray:
|
34 |
+
global embedding
|
35 |
image_input = Image.fromarray(image_input)
|
|
|
36 |
inputs = processor(image_input, input_points=points, return_tensors="pt").to(device)
|
37 |
+
if not isinstance(embedding, torch.Tensor):
|
38 |
+
embedding = model.get_image_embeddings(inputs["pixel_values"])
|
39 |
+
del inputs["pixel_values"]
|
40 |
+
|
41 |
+
outputs = model.forward(image_embeddings=embedding, **inputs)
|
42 |
masks = processor.image_processor.post_process_masks(
|
43 |
outputs.pred_masks.cpu(), inputs["original_sizes"].cpu(), inputs["reshaped_input_sizes"].cpu()
|
44 |
)
|
|
|
63 |
|
64 |
return pred_masks
|
65 |
|
66 |
+
def reset_embedding():
|
67 |
+
global embedding
|
68 |
+
embedding = None
|
69 |
|
70 |
with gr.Blocks() as demo:
|
71 |
gr.Markdown("# How to use")
|
|
|
81 |
image_button = gr.Button("Segment Image")
|
82 |
|
83 |
image_button.click(main_func, inputs=image_input, outputs=image_output)
|
84 |
+
image_input.upload(reset_embedding)
|
85 |
|
86 |
demo.launch()
|
requirements.txt
CHANGED
@@ -1,3 +1,4 @@
|
|
1 |
torch
|
2 |
git+https://github.com/huggingface/transformers
|
3 |
-
opencv-python
|
|
|
|
1 |
torch
|
2 |
git+https://github.com/huggingface/transformers
|
3 |
+
opencv-python
|
4 |
+
gradio --upgrade
|