Create app.py
Browse files
app.py
ADDED
@@ -0,0 +1,127 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import gradio as gr
|
2 |
+
import numpy as np
|
3 |
+
import torch
|
4 |
+
from PIL import Image
|
5 |
+
from transformers import SamModel, SamProcessor
|
6 |
+
from gradio_image_prompter import ImagePrompter
|
7 |
+
|
8 |
+
|
9 |
+
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
|
10 |
+
sam_model = SamModel.from_pretrained("facebook/sam-vit-huge").to(device)
|
11 |
+
sam_processor = SamProcessor.from_pretrained("facebook/sam-vit-huge")
|
12 |
+
slimsam_model = SamModel.from_pretrained("nielsr/slimsam-50-uniform").to(device)
|
13 |
+
slimsam_processor = SamProcessor.from_pretrained("nielsr/slimsam-50-uniform")
|
14 |
+
|
15 |
+
def sam_box_inference(image, model, x_min, y_min, x_max, y_max):
|
16 |
+
|
17 |
+
inputs = sam_processor(
|
18 |
+
Image.fromarray(image),
|
19 |
+
input_boxes=[[[[x_min, y_min, x_max, y_max]]]],
|
20 |
+
return_tensors="pt"
|
21 |
+
).to(device)
|
22 |
+
|
23 |
+
with torch.no_grad():
|
24 |
+
outputs = model(**inputs)
|
25 |
+
|
26 |
+
mask = sam_processor.image_processor.post_process_masks(
|
27 |
+
outputs.pred_masks.cpu(),
|
28 |
+
inputs["original_sizes"].cpu(),
|
29 |
+
inputs["reshaped_input_sizes"].cpu()
|
30 |
+
)[0][0][0].numpy()
|
31 |
+
mask = mask[np.newaxis, ...]
|
32 |
+
print(mask)
|
33 |
+
print(mask.shape)
|
34 |
+
return [(mask, "mask")]
|
35 |
+
|
36 |
+
|
37 |
+
def sam_point_inference(image, model, x, y):
|
38 |
+
inputs = sam_processor(
|
39 |
+
image,
|
40 |
+
input_points=[[[x, y]]],
|
41 |
+
return_tensors="pt").to(device)
|
42 |
+
|
43 |
+
with torch.no_grad():
|
44 |
+
outputs = sam_model(**inputs)
|
45 |
+
|
46 |
+
mask = sam_processor.post_process_masks(
|
47 |
+
outputs.pred_masks.cpu(),
|
48 |
+
inputs["original_sizes"].cpu(),
|
49 |
+
inputs["reshaped_input_sizes"].cpu()
|
50 |
+
)[0][0][0].numpy()
|
51 |
+
mask = mask[np.newaxis, ...]
|
52 |
+
print(type(mask))
|
53 |
+
print(mask.shape)
|
54 |
+
return [(mask, "mask")]
|
55 |
+
|
56 |
+
def infer_point(img):
|
57 |
+
if img is None:
|
58 |
+
gr.Error("Please upload an image and select a point.")
|
59 |
+
if img["background"] is None:
|
60 |
+
gr.Error("Please upload an image and select a point.")
|
61 |
+
# background (original image) layers[0] ( point prompt) composite (total image)
|
62 |
+
image = img["background"].convert("RGB")
|
63 |
+
point_prompt = img["layers"][0]
|
64 |
+
total_image = img["composite"]
|
65 |
+
img_arr = np.array(point_prompt)
|
66 |
+
if not np.any(img_arr):
|
67 |
+
gr.Error("Please select a point on top of the image.")
|
68 |
+
else:
|
69 |
+
nonzero_indices = np.nonzero(img_arr)
|
70 |
+
img_arr = np.array(point_prompt)
|
71 |
+
nonzero_indices = np.nonzero(img_arr)
|
72 |
+
center_x = int(np.mean(nonzero_indices[1]))
|
73 |
+
center_y = int(np.mean(nonzero_indices[0]))
|
74 |
+
print("Point inference returned.")
|
75 |
+
return ((image, sam_point_inference(image, slimsam_model, center_x, center_y)),
|
76 |
+
(image, sam_point_inference(image, sam_model, center_x, center_y)))
|
77 |
+
|
78 |
+
def infer_box(prompts):
|
79 |
+
# background (original image) layers[0] ( point prompt) composite (total image)
|
80 |
+
image = prompts["image"]
|
81 |
+
if image is None:
|
82 |
+
gr.Error("Please upload an image and draw a box before submitting")
|
83 |
+
points = prompts["points"][0]
|
84 |
+
if points is None:
|
85 |
+
gr.Error("Please draw a box before submitting.")
|
86 |
+
print(points)
|
87 |
+
|
88 |
+
# x_min = points[0] x_max = points[3] y_min = points[1] y_max = points[4]
|
89 |
+
return ((image, sam_box_inference(image, slimsam_model, points[0], points[1], points[3], points[4])),
|
90 |
+
(image, sam_box_inference(image, sam_model, points[0], points[1], points[3], points[4])))
|
91 |
+
with gr.Blocks(title="SlimSAM") as demo:
|
92 |
+
gr.Markdown("# SlimSAM")
|
93 |
+
gr.Markdown("SlimSAM is the pruned-distilled version of SAM that is smaller.")
|
94 |
+
gr.Markdown("In this demo, you can compare SlimSAM and SAM outputs in point and box prompts.")
|
95 |
+
|
96 |
+
with gr.Tab("Box Prompt"):
|
97 |
+
with gr.Row():
|
98 |
+
with gr.Column(scale=1):
|
99 |
+
# Title
|
100 |
+
gr.Markdown("Box Prompting")
|
101 |
+
with gr.Row():
|
102 |
+
with gr.Column():
|
103 |
+
im = ImagePrompter()
|
104 |
+
btn = gr.Button("Submit")
|
105 |
+
with gr.Column():
|
106 |
+
output_box_slimsam = gr.AnnotatedImage(label="SlimSAM Output")
|
107 |
+
output_box_sam = gr.AnnotatedImage(label="SAM Output")
|
108 |
+
|
109 |
+
|
110 |
+
btn.click(infer_box, inputs=im, outputs=[output_box_slimsam, output_box_sam])
|
111 |
+
|
112 |
+
with gr.Tab("Point Prompt"):
|
113 |
+
with gr.Row():
|
114 |
+
with gr.Column(scale=1):
|
115 |
+
# Title
|
116 |
+
gr.Markdown("Point Prompting")
|
117 |
+
with gr.Row():
|
118 |
+
with gr.Column():
|
119 |
+
im = gr.ImageEditor(
|
120 |
+
type="pil",
|
121 |
+
)
|
122 |
+
with gr.Column():
|
123 |
+
output_slimsam = gr.AnnotatedImage(label="SlimSAM Output")
|
124 |
+
output_sam = gr.AnnotatedImage(label="SAM Output")
|
125 |
+
|
126 |
+
im.change(infer_point, inputs=im, outputs=[output_slimsam, output_sam])
|
127 |
+
demo.launch(debug=True)
|