Spaces:
Running
on
T4
Running
on
T4
Create app.py
Browse files
app.py
ADDED
@@ -0,0 +1,128 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
from optimum.onnxruntime import ORTModelForVision2Seq
|
2 |
+
from transformers import TrOCRProcessor
|
3 |
+
from ultralytics import YOLO
|
4 |
+
import gradio as gr
|
5 |
+
import numpy as np
|
6 |
+
import onnxruntime
|
7 |
+
import time
|
8 |
+
|
9 |
+
from plotting_functions import PlotHTR
|
10 |
+
from segment_image import SegmentImage
|
11 |
+
from onnx_text_recognition import TextRecognition
|
12 |
+
|
13 |
+
|
14 |
+
LINE_MODEL_PATH = "Kansallisarkisto/multicentury-textline-detection"
|
15 |
+
REGION_MODEL_PATH = "Kansallisarkisto/court-records-region-detection"
|
16 |
+
TROCR_PROCESSOR_PATH = "Kansallisarkisto/multicentury-htr-model-onnx/202405_processor/"
|
17 |
+
TROCR_MODEL_PATH = "Kansallisarkisto/multicentury-htr-model-onnx/202405_onnx/"
|
18 |
+
|
19 |
+
|
20 |
+
def get_segmenter():
|
21 |
+
"""Initialize segmentation class."""
|
22 |
+
try:
|
23 |
+
segmenter = SegmentImage(line_model_path=LINE_MODEL_PATH,
|
24 |
+
device='cpu',
|
25 |
+
line_iou=0.3,
|
26 |
+
region_iou=0.5,
|
27 |
+
line_overlap=0.5,
|
28 |
+
line_nms_iou=0.7,
|
29 |
+
region_nms_iou=0.3,
|
30 |
+
line_conf_threshold=0.25,
|
31 |
+
region_conf_threshold=0.5,
|
32 |
+
region_model_path=REGION_MODEL_PATH,
|
33 |
+
order_regions=True,
|
34 |
+
region_half_precision=False,
|
35 |
+
line_half_precision=False)
|
36 |
+
return segmenter
|
37 |
+
except Exception as e:
|
38 |
+
print('Failed to initialize SegmentImage class: %s' % e)
|
39 |
+
|
40 |
+
def get_recognizer():
|
41 |
+
"""Initialize text recognition class."""
|
42 |
+
try:
|
43 |
+
recognizer = TextRecognition(
|
44 |
+
processor_path = TROCR_PROCESSOR_PATH,
|
45 |
+
model_path = TROCR_MODEL_PATH,
|
46 |
+
device = 'cpu',
|
47 |
+
half_precision = True,
|
48 |
+
line_threshold = 100
|
49 |
+
)
|
50 |
+
return recognizer
|
51 |
+
except Exception as e:
|
52 |
+
print('Failed to initialize TextRecognition class: %s' % e)
|
53 |
+
|
54 |
+
segmenter = get_segmenter()
|
55 |
+
recognizer = get_recognizer()
|
56 |
+
plotter = PlotHTR()
|
57 |
+
|
58 |
+
color_codes = """**Text region type:** <br>
|
59 |
+
Paragraph 
|
60 |
+
Marginalia 
|
61 |
+
Page number """
|
62 |
+
|
63 |
+
def merge_lines(segment_predictions):
|
64 |
+
img_lines = []
|
65 |
+
for region in segment_predictions:
|
66 |
+
img_lines += region['lines']
|
67 |
+
return img_lines
|
68 |
+
|
69 |
+
def get_text_predictions(image, segment_predictions, recognizer):
|
70 |
+
"""Collects text prediction data into dicts based on detected text regions."""
|
71 |
+
img_lines = merge_lines(segment_predictions)
|
72 |
+
height, width = segment_predictions[0]['img_shape']
|
73 |
+
# Process all lines of an image
|
74 |
+
texts = recognizer.process_lines(img_lines, image, height, width)
|
75 |
+
return texts
|
76 |
+
|
77 |
+
# Run demo code
|
78 |
+
with gr.Blocks(theme=gr.themes.Monochrome(), title="HTR demo") as demo:
|
79 |
+
gr.Markdown("# HTR demo")
|
80 |
+
with gr.Tab("Text content"):
|
81 |
+
with gr.Row():
|
82 |
+
input_img = gr.Image(label="Input image", type="pil")
|
83 |
+
textbox = gr.Textbox(label="Predicted text content", lines=10)
|
84 |
+
button = gr.Button("Process image")
|
85 |
+
processing_time = gr.Markdown()
|
86 |
+
with gr.Tab("Text regions"):
|
87 |
+
region_img = gr.Image(label="Predicted text regions", type="numpy")
|
88 |
+
gr.Markdown(color_codes)
|
89 |
+
with gr.Tab("Text lines"):
|
90 |
+
line_img = gr.Image(label="Predicted text lines", type="numpy")
|
91 |
+
gr.Markdown(color_codes)
|
92 |
+
|
93 |
+
def run_pipeline(image):
|
94 |
+
# Predict region and line segments
|
95 |
+
start = time.time()
|
96 |
+
segment_predictions = segmenter.get_segmentation(image)
|
97 |
+
if segment_predictions:
|
98 |
+
region_plot = plotter.plot_regions(segment_predictions, image)
|
99 |
+
line_plot = plotter.plot_lines(segment_predictions, image)
|
100 |
+
text_predictions = get_text_predictions(np.array(image), segment_predictions, recognizer)
|
101 |
+
text = "\n".join(text_predictions)
|
102 |
+
end = time.time()
|
103 |
+
proc_time = end - start
|
104 |
+
proc_time_str = f"Processing time: {proc_time:.4f}s"
|
105 |
+
return {
|
106 |
+
region_img: region_plot,
|
107 |
+
line_img: line_plot,
|
108 |
+
textbox: text,
|
109 |
+
processing_time: proc_time_str
|
110 |
+
}
|
111 |
+
else:
|
112 |
+
end = time.time()
|
113 |
+
proc_time = end - start
|
114 |
+
proc_time_str = f"Processing time: {proc_time:.4f}s"
|
115 |
+
return {
|
116 |
+
region_img: None,
|
117 |
+
line_img: None,
|
118 |
+
textbox: None,
|
119 |
+
processing_time: proc_time_str
|
120 |
+
}
|
121 |
+
|
122 |
+
|
123 |
+
button.click(fn=run_pipeline,
|
124 |
+
inputs=input_img,
|
125 |
+
outputs=[region_img, line_img, textbox, processing_time])
|
126 |
+
|
127 |
+
if __name__ == "__main__":
|
128 |
+
demo.launch()
|