Spaces:
Sleeping
Sleeping
wjbmattingly
commited on
Commit
•
05d0d72
1
Parent(s):
318c115
Update app.py
Browse files
app.py
CHANGED
@@ -14,48 +14,15 @@ import os
|
|
14 |
|
15 |
subprocess.run('pip install flash-attn --no-build-isolation', env={'FLASH_ATTENTION_SKIP_CUDA_BUILD': "TRUE"}, shell=True)
|
16 |
|
17 |
-
model_dir = 'medieval-data/florence2-medieval-bbox-line-detection'
|
18 |
model_dir = "medieval-data/florence2-medieval-bbox-zone-detection"
|
19 |
|
20 |
-
|
21 |
-
|
22 |
-
|
23 |
-
|
24 |
-
|
25 |
-
|
26 |
-
|
27 |
-
|
28 |
-
with patch("transformers.dynamic_module_utils.get_imports", fixed_get_imports):
|
29 |
-
# Load the configuration
|
30 |
-
config = AutoConfig.from_pretrained(model_dir, trust_remote_code=True)
|
31 |
-
|
32 |
-
# Modify the vision configuration
|
33 |
-
if hasattr(config, 'vision_config'):
|
34 |
-
config.vision_config.model_type = 'davit'
|
35 |
-
|
36 |
-
print("Modified vision configuration:")
|
37 |
-
print(config.vision_config)
|
38 |
-
|
39 |
-
# Try to load the model with the modified configuration
|
40 |
-
try:
|
41 |
-
model = AutoModelForCausalLM.from_pretrained(
|
42 |
-
model_dir,
|
43 |
-
config=config,
|
44 |
-
trust_remote_code=True
|
45 |
-
)
|
46 |
-
print("Model loaded successfully!")
|
47 |
-
except Exception as e:
|
48 |
-
print(f"Failed to load model: {str(e)}")
|
49 |
-
|
50 |
-
# Load the processor without specifying a revision
|
51 |
-
try:
|
52 |
-
processor = AutoProcessor.from_pretrained(
|
53 |
-
model_dir,
|
54 |
-
trust_remote_code=True
|
55 |
-
)
|
56 |
-
print("Processor loaded successfully!")
|
57 |
-
except Exception as e:
|
58 |
-
print(f"Failed to load processor: {str(e)}")
|
59 |
TITLE = "# [Florence-2- Medieval Manuscript Layout Parsing Demo](https://huggingface.co/medieval-data/florence2-medieval-bbox-zone-detection)"
|
60 |
DESCRIPTION = "The demo for Florence-2 fine-tuned on CATMuS Segmentation Dataset. This app has two models: one for line detection and one for zone detection."
|
61 |
|
@@ -63,7 +30,7 @@ DESCRIPTION = "The demo for Florence-2 fine-tuned on CATMuS Segmentation Dataset
|
|
63 |
colormap = plt.cm.get_cmap('tab20')
|
64 |
|
65 |
@spaces.GPU
|
66 |
-
def process_image(image
|
67 |
max_size = 1000
|
68 |
prompt = "<OD>"
|
69 |
|
@@ -111,14 +78,14 @@ def visualize_bboxes(result, image):
|
|
111 |
plt.axis('off')
|
112 |
return fig
|
113 |
|
114 |
-
def run_example(image
|
115 |
if isinstance(image, str): # If image is a URL
|
116 |
response = requests.get(image)
|
117 |
image = Image.open(BytesIO(response.content))
|
118 |
elif isinstance(image, np.ndarray): # If image is a numpy array
|
119 |
image = Image.fromarray(image)
|
120 |
|
121 |
-
result, processed_image = process_image(image
|
122 |
fig = visualize_bboxes(result, processed_image)
|
123 |
|
124 |
# Convert matplotlib figure to image
|
@@ -128,15 +95,13 @@ def run_example(image, text_input=None):
|
|
128 |
output_image = Image.open(img_buf)
|
129 |
|
130 |
return output_image
|
131 |
-
|
132 |
css = """
|
133 |
#output {
|
134 |
-
height:
|
135 |
overflow: auto;
|
136 |
border: 1px solid #ccc;
|
137 |
}
|
138 |
"""
|
139 |
-
|
140 |
with gr.Blocks(css=css) as demo:
|
141 |
gr.Markdown(TITLE)
|
142 |
gr.Markdown(DESCRIPTION)
|
@@ -144,23 +109,19 @@ with gr.Blocks(css=css) as demo:
|
|
144 |
with gr.Row():
|
145 |
with gr.Column():
|
146 |
input_img = gr.Image(label="Input Picture")
|
147 |
-
text_input = gr.Textbox(label="Text Input (optional)")
|
148 |
submit_btn = gr.Button(value="Submit")
|
149 |
with gr.Column():
|
150 |
output_img = gr.Image(label="Output Image with Bounding Boxes")
|
151 |
gr.Examples(
|
152 |
examples=[
|
153 |
-
["https://huggingface.co/datasets/CATMuS/medieval-segmentation/resolve/main/data/dev/london-british-library-egerton-821/page-002-of-004.jpg"
|
154 |
-
["https://huggingface.co/datasets/CATMuS/medieval-segmentation/resolve/main/data/dev/paris-bnf-lat-12449/page-002-of-003.jpg", None],
|
155 |
-
["https://huggingface.co/datasets/CATMuS/medieval-segmentation/resolve/main/data/dev/paris-bnf-nal-1909/page-009-of-012.jpg", None],
|
156 |
-
["https://huggingface.co/datasets/CATMuS/medieval-segmentation/resolve/main/data/test/paris-bnf-fr-574/page-001-of-003.jpg", None]
|
157 |
],
|
158 |
-
inputs=[input_img
|
159 |
outputs=[output_img],
|
160 |
fn=run_example,
|
161 |
cache_examples=True,
|
162 |
label='Try the examples below'
|
163 |
)
|
164 |
-
submit_btn.click(run_example, [input_img
|
165 |
|
166 |
demo.launch(debug=True)
|
|
|
14 |
|
15 |
subprocess.run('pip install flash-attn --no-build-isolation', env={'FLASH_ATTENTION_SKIP_CUDA_BUILD': "TRUE"}, shell=True)
|
16 |
|
|
|
17 |
model_dir = "medieval-data/florence2-medieval-bbox-zone-detection"
|
18 |
|
19 |
+
# Load the configuration
|
20 |
+
config = AutoConfig.from_pretrained(model_dir, trust_remote_code=True)
|
21 |
+
model = AutoModelForCausalLM.from_pretrained(
|
22 |
+
model_dir,
|
23 |
+
trust_remote_code=True
|
24 |
+
)
|
25 |
+
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
26 |
TITLE = "# [Florence-2- Medieval Manuscript Layout Parsing Demo](https://huggingface.co/medieval-data/florence2-medieval-bbox-zone-detection)"
|
27 |
DESCRIPTION = "The demo for Florence-2 fine-tuned on CATMuS Segmentation Dataset. This app has two models: one for line detection and one for zone detection."
|
28 |
|
|
|
30 |
colormap = plt.cm.get_cmap('tab20')
|
31 |
|
32 |
@spaces.GPU
|
33 |
+
def process_image(image):
|
34 |
max_size = 1000
|
35 |
prompt = "<OD>"
|
36 |
|
|
|
78 |
plt.axis('off')
|
79 |
return fig
|
80 |
|
81 |
+
def run_example(image):
|
82 |
if isinstance(image, str): # If image is a URL
|
83 |
response = requests.get(image)
|
84 |
image = Image.open(BytesIO(response.content))
|
85 |
elif isinstance(image, np.ndarray): # If image is a numpy array
|
86 |
image = Image.fromarray(image)
|
87 |
|
88 |
+
result, processed_image = process_image(image)
|
89 |
fig = visualize_bboxes(result, processed_image)
|
90 |
|
91 |
# Convert matplotlib figure to image
|
|
|
95 |
output_image = Image.open(img_buf)
|
96 |
|
97 |
return output_image
|
|
|
98 |
css = """
|
99 |
#output {
|
100 |
+
height: 1000px;
|
101 |
overflow: auto;
|
102 |
border: 1px solid #ccc;
|
103 |
}
|
104 |
"""
|
|
|
105 |
with gr.Blocks(css=css) as demo:
|
106 |
gr.Markdown(TITLE)
|
107 |
gr.Markdown(DESCRIPTION)
|
|
|
109 |
with gr.Row():
|
110 |
with gr.Column():
|
111 |
input_img = gr.Image(label="Input Picture")
|
|
|
112 |
submit_btn = gr.Button(value="Submit")
|
113 |
with gr.Column():
|
114 |
output_img = gr.Image(label="Output Image with Bounding Boxes")
|
115 |
gr.Examples(
|
116 |
examples=[
|
117 |
+
["https://huggingface.co/datasets/CATMuS/medieval-segmentation/resolve/main/data/dev/london-british-library-egerton-821/page-002-of-004.jpg"],
|
|
|
|
|
|
|
118 |
],
|
119 |
+
inputs=[input_img],
|
120 |
outputs=[output_img],
|
121 |
fn=run_example,
|
122 |
cache_examples=True,
|
123 |
label='Try the examples below'
|
124 |
)
|
125 |
+
submit_btn.click(run_example, [input_img], [output_img])
|
126 |
|
127 |
demo.launch(debug=True)
|