Spaces:
Sleeping
Sleeping
File size: 16,362 Bytes
34b98f1 b76cf08 34b98f1 14d4a0b 3b057c5 60ad418 3b057c5 14d4a0b 3b057c5 b76cf08 14d4a0b b76cf08 3b057c5 b76cf08 3b057c5 b76cf08 3b057c5 b76cf08 7263d32 b76cf08 14d4a0b b76cf08 14d4a0b b76cf08 14d4a0b b76cf08 3b057c5 b76cf08 60ad418 b76cf08 3b057c5 b76cf08 3b057c5 7263d32 3b057c5 b76cf08 3b057c5 b76cf08 7263d32 b76cf08 3b057c5 7263d32 3b057c5 b76cf08 3b057c5 7263d32 3b057c5 7263d32 3b057c5 b76cf08 3b057c5 7263d32 3b057c5 b76cf08 3b057c5 7263d32 3b057c5 b76cf08 3b057c5 14d4a0b b76cf08 14d4a0b b76cf08 7263d32 3b057c5 b76cf08 3b057c5 7263d32 3b057c5 7263d32 3b057c5 b76cf08 3b057c5 7263d32 14d4a0b b76cf08 14d4a0b 3b057c5 b76cf08 7263d32 3b057c5 7263d32 3b057c5 b76cf08 3b057c5 7263d32 3b057c5 14d4a0b 3b057c5 14d4a0b 3b057c5 7263d32 3b057c5 14d4a0b 3b057c5 60ad418 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 |
import os
import shutil
from difflib import Differ
import evaluate
import gradio as gr
from helper.examples.examples import DemoImages
from helper.utils import TrafficDataHandler
from src.htr_pipeline.gradio_backend import CustomTrack, SingletonModelLoader
model_loader = SingletonModelLoader()
custom_track = CustomTrack(model_loader)
images_for_demo = DemoImages()
cer_metric = evaluate.load("cer")
with gr.Blocks() as stepwise_htr_tool_tab:
with gr.Tabs():
with gr.Tab("1. Region segmentation"):
with gr.Row():
with gr.Column(scale=1):
vis_data_folder_placeholder = gr.Markdown(visible=False)
name_files_placeholder = gr.Markdown(visible=False)
with gr.Group():
input_region_image = gr.Image(
label="Image to region segment",
# type="numpy",
tool="editor",
height=500,
)
with gr.Accordion("Settings", open=False):
with gr.Group():
reg_pred_score_threshold_slider = gr.Slider(
minimum=0.4,
maximum=1,
value=0.5,
step=0.05,
label="P-threshold",
info="""Filter the confidence score for a prediction score to be considered""",
)
reg_containments_threshold_slider = gr.Slider(
minimum=0,
maximum=1,
value=0.5,
step=0.05,
label="C-threshold",
info="""The minimum required overlap or similarity
for a detected region or object to be considered valid""",
)
region_segment_model_dropdown = gr.Dropdown(
choices=["Riksarkivet/rtm_region"],
value="Riksarkivet/rtm_region",
label="Region segmentation model",
info="More models will be added",
)
with gr.Row():
clear_button = gr.Button("Clear", variant="secondary", elem_id="clear_button")
region_segment_button = gr.Button(
"Run",
variant="primary",
elem_id="region_segment_button",
)
region_segment_button_var = gr.State(value="region_segment_button")
with gr.Column(scale=2):
with gr.Box():
with gr.Row():
with gr.Column(scale=2):
gr.Examples(
examples=images_for_demo.examples_list,
inputs=[name_files_placeholder, input_region_image],
label="Example images",
examples_per_page=5,
)
with gr.Column(scale=3):
output_region_image = gr.Image(label="Segmented regions", type="numpy", height=600)
##############################################
with gr.Tab("2. Line segmentation"):
image_placeholder_lines = gr.Image(
label="Segmented lines",
# type="numpy",
interactive="False",
visible=True,
height=600,
)
with gr.Row(visible=False) as control_line_segment:
with gr.Column(scale=2):
with gr.Group():
with gr.Box():
regions_cropped_gallery = gr.Gallery(
label="Segmented regions",
elem_id="gallery",
columns=[2],
rows=[2],
# object_fit="contain",
height=450,
preview=True,
container=False,
)
input_region_from_gallery = gr.Image(
label="Region segmentation to line segment", interactive="False", visible=False, height=400
)
with gr.Row():
with gr.Accordion("Settings", open=False):
with gr.Row():
line_pred_score_threshold_slider = gr.Slider(
minimum=0.3,
maximum=1,
value=0.4,
step=0.05,
label="Pred_score threshold",
info="""Filter the confidence score for a prediction score to be considered""",
)
line_containments_threshold_slider = gr.Slider(
minimum=0,
maximum=1,
value=0.5,
step=0.05,
label="Containments threshold",
info="""The minimum required overlap or similarity
for a detected region or object to be considered valid""",
)
with gr.Row(equal_height=False):
line_segment_model_dropdown = gr.Dropdown(
choices=["Riksarkivet/rtmdet_lines"],
value="Riksarkivet/rtmdet_lines",
label="Line segment model",
info="More models will be added",
)
with gr.Row():
clear_line_segment_button = gr.Button(
" ",
variant="Secondary",
# elem_id="center_button",
scale=1,
)
line_segment_button = gr.Button(
"Run",
variant="primary",
# elem_id="center_button",
scale=1,
)
with gr.Column(scale=3):
# gr.Markdown("""lorem ipsum""")
output_line_from_region = gr.Image(
label="Segmented lines", type="numpy", interactive="False", height=600
)
###############################################
with gr.Tab("3. Text recognition"):
image_placeholder_htr = gr.Image(
label="Transcribed lines",
# type="numpy",
interactive="False",
visible=True,
height=600,
)
with gr.Row(visible=False) as control_htr:
inputs_lines_to_transcribe = gr.Variable()
with gr.Column(scale=2):
with gr.Group():
image_inputs_lines_to_transcribe = gr.Image(
label="Transcribed lines", type="numpy", interactive="False", visible=False, height=470
)
with gr.Row():
with gr.Accordion("Settings", open=False):
transcriber_model = gr.Dropdown(
choices=["Riksarkivet/satrn_htr", "microsoft/trocr-base-handwritten"],
value="Riksarkivet/satrn_htr",
label="Text recognition model",
info="More models will be added",
)
gr.Slider(
value=0.6,
minimum=0.5,
maximum=1,
label="HTR threshold",
info="Prediction score threshold for transcribed lines",
scale=1,
)
with gr.Row():
clear_transcribe_button = gr.Button(" ", variant="Secondary", visible=True, scale=1)
transcribe_button = gr.Button("Run", variant="primary", visible=True, scale=1)
with gr.Column(scale=3):
with gr.Row():
transcribed_text = gr.Textbox(
label="Transcribed text",
info="Transcribed text is being streamed back from the Text recognition model",
lines=26,
value="",
show_copy_button=True,
)
#####################################
with gr.Tab("4. Explore results"):
image_placeholder_explore_results = gr.Image(
label="Cropped transcribed lines",
# type="numpy",
interactive="False",
visible=True,
height=600,
)
with gr.Row(visible=False, equal_height=False) as control_results_transcribe:
with gr.Column(scale=1, visible=True):
with gr.Group():
with gr.Box():
temp_gallery_input = gr.Variable()
gallery_inputs_lines_to_transcribe = gr.Gallery(
label="Cropped transcribed lines",
elem_id="gallery_lines",
columns=[3],
rows=[3],
# object_fit="contain",
height=150,
preview=True,
container=False,
)
with gr.Row():
dataframe_text_index = gr.Textbox(
label="Text from DataFrame selection",
placeholder="Select row from the DataFrame.",
interactive=False,
)
with gr.Row():
gt_text_index = gr.Textbox(
label="Ground Truth",
placeholder="Provide the ground truth, if available.",
interactive=True,
)
with gr.Row():
diff_token_output = gr.HighlightedText(
label="Text diff",
combine_adjacent=True,
show_legend=True,
color_map={"+": "red", "-": "green"},
)
with gr.Row(equal_height=False):
cer_output = gr.Textbox(label="CER:")
calc_cer_button = gr.Button("Calculate CER", variant="primary", visible=True)
with gr.Column(scale=1, visible=True):
mapping_dict = gr.Variable()
transcribed_text_df_finish = gr.Dataframe(
headers=["Transcribed text", "Pred score"],
max_rows=14,
col_count=(2, "fixed"),
wrap=True,
interactive=False,
overflow_row_behaviour="paginate",
height=600,
)
# custom track
def diff_texts(text1, text2):
d = Differ()
return [(token[2:], token[0] if token[0] != " " else None) for token in d.compare(text1, text2)]
def compute_cer(dataframe_text_index, gt_text_index):
if gt_text_index is not None and gt_text_index.strip() != "":
return cer_metric.compute(predictions=[dataframe_text_index], references=[gt_text_index])
else:
return "Ground truth not provided"
calc_cer_button.click(compute_cer, inputs=[dataframe_text_index, gt_text_index], outputs=cer_output)
calc_cer_button.click(diff_texts, inputs=[dataframe_text_index, gt_text_index], outputs=[diff_token_output])
region_segment_button.click(
custom_track.region_segment,
inputs=[input_region_image, reg_pred_score_threshold_slider, reg_containments_threshold_slider],
outputs=[output_region_image, regions_cropped_gallery, image_placeholder_lines, control_line_segment],
)
regions_cropped_gallery.select(
custom_track.get_select_index_image, regions_cropped_gallery, input_region_from_gallery
)
transcribed_text_df_finish.select(
fn=custom_track.get_select_index_df,
inputs=[transcribed_text_df_finish, mapping_dict],
outputs=[gallery_inputs_lines_to_transcribe, dataframe_text_index],
)
line_segment_button.click(
custom_track.line_segment,
inputs=[input_region_from_gallery, line_pred_score_threshold_slider, line_containments_threshold_slider],
outputs=[
output_line_from_region,
image_inputs_lines_to_transcribe,
inputs_lines_to_transcribe,
gallery_inputs_lines_to_transcribe,
temp_gallery_input,
# Hide
transcribe_button,
image_inputs_lines_to_transcribe,
image_placeholder_htr,
control_htr,
],
)
transcribe_button.click(
custom_track.transcribe_text,
inputs=[inputs_lines_to_transcribe],
outputs=[
transcribed_text,
transcribed_text_df_finish,
mapping_dict,
# Hide
control_results_transcribe,
image_placeholder_explore_results,
],
)
clear_button.click(
lambda: (
(shutil.rmtree("./vis_data") if os.path.exists("./vis_data") else None, None)[1],
None,
None,
None,
gr.update(visible=False),
None,
None,
None,
gr.update(visible=False),
gr.update(visible=False),
gr.update(visible=True),
None,
gr.update(visible=False),
gr.update(visible=False),
gr.update(visible=True),
gr.update(visible=True),
),
inputs=[],
outputs=[
vis_data_folder_placeholder,
input_region_image,
regions_cropped_gallery,
input_region_from_gallery,
control_line_segment,
output_line_from_region,
inputs_lines_to_transcribe,
transcribed_text,
control_htr,
inputs_lines_to_transcribe,
image_placeholder_htr,
output_region_image,
image_inputs_lines_to_transcribe,
control_results_transcribe,
image_placeholder_explore_results,
image_placeholder_lines,
],
)
region_segment_button.click(fn=TrafficDataHandler.store_metric_data, inputs=region_segment_button_var)
|