Spaces:
Running
on
Zero
Running
on
Zero
Commit
·
2c499db
1
Parent(s):
275bb85
Refactor XML parsing functions to support both ALTO and PAGE formats, enhancing error handling and output consistency
Browse files
app.py
CHANGED
@@ -23,18 +23,65 @@ HF_PIPE = pipeline("image-text-to-text", model=HF_MODEL, processor=HF_PROCESSOR)
|
|
23 |
|
24 |
# --- Helper Functions ---
|
25 |
|
26 |
-
def
|
27 |
"""
|
28 |
-
Dynamically gets the
|
|
|
29 |
"""
|
30 |
try:
|
31 |
tree = ET.parse(xml_file_path)
|
32 |
root = tree.getroot()
|
33 |
if '}' in root.tag:
|
34 |
-
|
|
|
|
|
|
|
|
|
|
|
35 |
except ET.ParseError:
|
36 |
print(f"Error parsing XML to find namespace: {xml_file_path}")
|
37 |
-
return ''
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
38 |
|
39 |
def parse_alto_xml_for_text(xml_file_path):
|
40 |
"""
|
@@ -48,7 +95,7 @@ def parse_alto_xml_for_text(xml_file_path):
|
|
48 |
return "Error: XML file not provided or does not exist."
|
49 |
|
50 |
try:
|
51 |
-
ns_prefix =
|
52 |
tree = ET.parse(xml_file_path)
|
53 |
root = tree.getroot()
|
54 |
|
@@ -68,6 +115,26 @@ def parse_alto_xml_for_text(xml_file_path):
|
|
68 |
except Exception as e:
|
69 |
return f"An unexpected error occurred during XML parsing: {e}"
|
70 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
71 |
@spaces.GPU
|
72 |
def predict(pil_image):
|
73 |
"""Performs OCR prediction using the Hugging Face model."""
|
@@ -148,10 +215,10 @@ def process_files(image_path, xml_path):
|
|
148 |
"""
|
149 |
Main function for the Gradio interface.
|
150 |
Processes the image for display, runs OCR (Hugging Face model),
|
151 |
-
and parses
|
152 |
"""
|
153 |
img_to_display = None
|
154 |
-
|
155 |
hf_ocr_text_output = "Image not provided or OCR not run."
|
156 |
|
157 |
if image_path:
|
@@ -164,19 +231,17 @@ def process_files(image_path, xml_path):
|
|
164 |
else:
|
165 |
hf_ocr_text_output = "Please upload an image to perform OCR."
|
166 |
|
167 |
-
|
168 |
if xml_path:
|
169 |
-
|
170 |
else:
|
171 |
-
|
172 |
|
173 |
# If only XML is provided without an image
|
174 |
if not image_path and xml_path:
|
175 |
img_to_display = None # No image to display
|
176 |
hf_ocr_text_output = "Upload an image to perform OCR."
|
177 |
|
178 |
-
|
179 |
-
return img_to_display, alto_text_output, hf_ocr_text_output
|
180 |
|
181 |
|
182 |
# --- Create Gradio App ---
|
@@ -185,13 +250,13 @@ with gr.Blocks(theme=gr.themes.Soft()) as demo:
|
|
185 |
gr.Markdown("# OCR Viewer and Extractor")
|
186 |
gr.Markdown(
|
187 |
"Upload an image to perform OCR using a Hugging Face model. "
|
188 |
-
"Optionally, upload its corresponding ALTO
|
189 |
)
|
190 |
|
191 |
with gr.Row():
|
192 |
with gr.Column(scale=1):
|
193 |
image_input = gr.File(label="Upload Image (PNG, JPG, etc.)", type="filepath")
|
194 |
-
xml_input = gr.File(label="Upload
|
195 |
submit_button = gr.Button("Process Image and XML", variant="primary")
|
196 |
|
197 |
with gr.Row():
|
@@ -204,8 +269,8 @@ with gr.Blocks(theme=gr.themes.Soft()) as demo:
|
|
204 |
interactive=False,
|
205 |
show_copy_button=True
|
206 |
)
|
207 |
-
|
208 |
-
label="Text from
|
209 |
lines=15,
|
210 |
interactive=False,
|
211 |
show_copy_button=True
|
@@ -214,7 +279,7 @@ with gr.Blocks(theme=gr.themes.Soft()) as demo:
|
|
214 |
submit_button.click(
|
215 |
fn=process_files,
|
216 |
inputs=[image_input, xml_input],
|
217 |
-
outputs=[output_image_display,
|
218 |
)
|
219 |
|
220 |
gr.Markdown("---")
|
|
|
23 |
|
24 |
# --- Helper Functions ---
|
25 |
|
26 |
+
def get_xml_namespace(xml_file_path):
|
27 |
"""
|
28 |
+
Dynamically gets the namespace from the XML file.
|
29 |
+
Returns both the namespace and the format type (ALTO or PAGE).
|
30 |
"""
|
31 |
try:
|
32 |
tree = ET.parse(xml_file_path)
|
33 |
root = tree.getroot()
|
34 |
if '}' in root.tag:
|
35 |
+
ns = root.tag.split('}')[0] + '}'
|
36 |
+
# Determine format based on root element
|
37 |
+
if 'PcGts' in root.tag:
|
38 |
+
return ns, 'PAGE'
|
39 |
+
elif 'alto' in root.tag.lower():
|
40 |
+
return ns, 'ALTO'
|
41 |
except ET.ParseError:
|
42 |
print(f"Error parsing XML to find namespace: {xml_file_path}")
|
43 |
+
return '', 'UNKNOWN'
|
44 |
+
|
45 |
+
def parse_page_xml_for_text(xml_file_path):
|
46 |
+
"""
|
47 |
+
Parses a PAGE XML file to extract text content.
|
48 |
+
Returns:
|
49 |
+
- full_text (str): All extracted text concatenated.
|
50 |
+
"""
|
51 |
+
full_text_lines = []
|
52 |
+
|
53 |
+
if not xml_file_path or not os.path.exists(xml_file_path):
|
54 |
+
return "Error: XML file not provided or does not exist."
|
55 |
+
|
56 |
+
try:
|
57 |
+
ns_prefix, _ = get_xml_namespace(xml_file_path)
|
58 |
+
tree = ET.parse(xml_file_path)
|
59 |
+
root = tree.getroot()
|
60 |
+
|
61 |
+
# Find all TextLine elements
|
62 |
+
for text_line in root.findall(f'.//{ns_prefix}TextLine'):
|
63 |
+
# First try to get text from TextEquiv/Unicode
|
64 |
+
text_equiv = text_line.find(f'{ns_prefix}TextEquiv/{ns_prefix}Unicode')
|
65 |
+
if text_equiv is not None and text_equiv.text:
|
66 |
+
full_text_lines.append(text_equiv.text)
|
67 |
+
continue
|
68 |
+
|
69 |
+
# If no TextEquiv, try to get text from Word elements
|
70 |
+
line_text_parts = []
|
71 |
+
for word in text_line.findall(f'{ns_prefix}Word'):
|
72 |
+
word_text = word.find(f'{ns_prefix}TextEquiv/{ns_prefix}Unicode')
|
73 |
+
if word_text is not None and word_text.text:
|
74 |
+
line_text_parts.append(word_text.text)
|
75 |
+
|
76 |
+
if line_text_parts:
|
77 |
+
full_text_lines.append(" ".join(line_text_parts))
|
78 |
+
|
79 |
+
return "\n".join(full_text_lines)
|
80 |
+
|
81 |
+
except ET.ParseError as e:
|
82 |
+
return f"Error parsing XML: {e}"
|
83 |
+
except Exception as e:
|
84 |
+
return f"An unexpected error occurred during XML parsing: {e}"
|
85 |
|
86 |
def parse_alto_xml_for_text(xml_file_path):
|
87 |
"""
|
|
|
95 |
return "Error: XML file not provided or does not exist."
|
96 |
|
97 |
try:
|
98 |
+
ns_prefix, _ = get_xml_namespace(xml_file_path)
|
99 |
tree = ET.parse(xml_file_path)
|
100 |
root = tree.getroot()
|
101 |
|
|
|
115 |
except Exception as e:
|
116 |
return f"An unexpected error occurred during XML parsing: {e}"
|
117 |
|
118 |
+
def parse_xml_for_text(xml_file_path):
|
119 |
+
"""
|
120 |
+
Main function to parse XML files, automatically detecting the format.
|
121 |
+
"""
|
122 |
+
if not xml_file_path or not os.path.exists(xml_file_path):
|
123 |
+
return "Error: XML file not provided or does not exist."
|
124 |
+
|
125 |
+
try:
|
126 |
+
_, xml_format = get_xml_namespace(xml_file_path)
|
127 |
+
|
128 |
+
if xml_format == 'PAGE':
|
129 |
+
return parse_page_xml_for_text(xml_file_path)
|
130 |
+
elif xml_format == 'ALTO':
|
131 |
+
return parse_alto_xml_for_text(xml_file_path)
|
132 |
+
else:
|
133 |
+
return f"Error: Unsupported XML format. Expected ALTO or PAGE XML."
|
134 |
+
|
135 |
+
except Exception as e:
|
136 |
+
return f"Error determining XML format: {str(e)}"
|
137 |
+
|
138 |
@spaces.GPU
|
139 |
def predict(pil_image):
|
140 |
"""Performs OCR prediction using the Hugging Face model."""
|
|
|
215 |
"""
|
216 |
Main function for the Gradio interface.
|
217 |
Processes the image for display, runs OCR (Hugging Face model),
|
218 |
+
and parses XML if provided.
|
219 |
"""
|
220 |
img_to_display = None
|
221 |
+
xml_text_output = "XML not provided or not processed."
|
222 |
hf_ocr_text_output = "Image not provided or OCR not run."
|
223 |
|
224 |
if image_path:
|
|
|
231 |
else:
|
232 |
hf_ocr_text_output = "Please upload an image to perform OCR."
|
233 |
|
|
|
234 |
if xml_path:
|
235 |
+
xml_text_output = parse_xml_for_text(xml_path)
|
236 |
else:
|
237 |
+
xml_text_output = "No XML file uploaded."
|
238 |
|
239 |
# If only XML is provided without an image
|
240 |
if not image_path and xml_path:
|
241 |
img_to_display = None # No image to display
|
242 |
hf_ocr_text_output = "Upload an image to perform OCR."
|
243 |
|
244 |
+
return img_to_display, xml_text_output, hf_ocr_text_output
|
|
|
245 |
|
246 |
|
247 |
# --- Create Gradio App ---
|
|
|
250 |
gr.Markdown("# OCR Viewer and Extractor")
|
251 |
gr.Markdown(
|
252 |
"Upload an image to perform OCR using a Hugging Face model. "
|
253 |
+
"Optionally, upload its corresponding ALTO or PAGE XML file to compare the extracted text."
|
254 |
)
|
255 |
|
256 |
with gr.Row():
|
257 |
with gr.Column(scale=1):
|
258 |
image_input = gr.File(label="Upload Image (PNG, JPG, etc.)", type="filepath")
|
259 |
+
xml_input = gr.File(label="Upload XML File (Optional, ALTO or PAGE format)", type="filepath")
|
260 |
submit_button = gr.Button("Process Image and XML", variant="primary")
|
261 |
|
262 |
with gr.Row():
|
|
|
269 |
interactive=False,
|
270 |
show_copy_button=True
|
271 |
)
|
272 |
+
xml_output_textbox = gr.Textbox(
|
273 |
+
label="Text from XML",
|
274 |
lines=15,
|
275 |
interactive=False,
|
276 |
show_copy_button=True
|
|
|
279 |
submit_button.click(
|
280 |
fn=process_files,
|
281 |
inputs=[image_input, xml_input],
|
282 |
+
outputs=[output_image_display, xml_output_textbox, hf_ocr_output_textbox]
|
283 |
)
|
284 |
|
285 |
gr.Markdown("---")
|