Update src/utils.py
Browse files- src/utils.py +172 -172
src/utils.py
CHANGED
@@ -1,172 +1,172 @@
|
|
1 |
-
from PIL import ImageDraw
|
2 |
-
import numpy as np
|
3 |
-
import re
|
4 |
-
|
5 |
-
|
6 |
-
# Use a color map for bounding boxes
|
7 |
-
colormap = [
|
8 |
-
"#0000FF",
|
9 |
-
"#FFA500",
|
10 |
-
"#008000",
|
11 |
-
"#800080",
|
12 |
-
"#A52A2A",
|
13 |
-
"#FFC0CB",
|
14 |
-
"#808080",
|
15 |
-
"#808000",
|
16 |
-
"#00FFFF",
|
17 |
-
"#FF0000",
|
18 |
-
"#00FF00",
|
19 |
-
"#4B0082",
|
20 |
-
"#4B0082",
|
21 |
-
"#EE82EE",
|
22 |
-
"#00FFFF",
|
23 |
-
"#FF00FF",
|
24 |
-
"#FF7F50",
|
25 |
-
"#FFD700",
|
26 |
-
"#87CEEB",
|
27 |
-
]
|
28 |
-
|
29 |
-
|
30 |
-
# Text cleaning function
|
31 |
-
def clean_text(text):
|
32 |
-
"""
|
33 |
-
Cleans the given text by removing unwanted tokens, extra spaces,
|
34 |
-
and ensures proper spacing between words and after periods.
|
35 |
-
|
36 |
-
Args:
|
37 |
-
text (str): The input text to be cleaned.
|
38 |
-
|
39 |
-
Returns:
|
40 |
-
str: The cleaned and properly formatted text.
|
41 |
-
"""
|
42 |
-
|
43 |
-
# Remove unwanted tokens
|
44 |
-
text = text.replace("<pad>", "").replace("</s>", "").strip()
|
45 |
-
|
46 |
-
# Split the text into lines and clean each line
|
47 |
-
lines = text.split("\n")
|
48 |
-
cleaned_lines = [line.strip() for line in lines if line.strip()]
|
49 |
-
|
50 |
-
# Join the cleaned lines into a single string with a space between each line
|
51 |
-
cleaned_text = " ".join(cleaned_lines)
|
52 |
-
|
53 |
-
# Ensure proper spacing between words and after periods using regex
|
54 |
-
cleaned_text = re.sub(
|
55 |
-
r"\s+", " ", cleaned_text
|
56 |
-
) # Replace multiple spaces with a single space
|
57 |
-
cleaned_text = re.sub(
|
58 |
-
r"(?<=[.])(?=[^\s])", r" ", cleaned_text
|
59 |
-
) # Add space after a period if not followed by a space
|
60 |
-
|
61 |
-
# Return the cleaned text
|
62 |
-
return cleaned_text
|
63 |
-
|
64 |
-
|
65 |
-
# Convert hex color to RGBA with the given alpha
|
66 |
-
def hex_to_rgba(hex_color, alpha):
|
67 |
-
"""
|
68 |
-
Convert a hexadecimal color code to RGBA format.
|
69 |
-
|
70 |
-
Args:
|
71 |
-
hex_color (str): The hexadecimal color code (e.g., "#FF0000").
|
72 |
-
alpha (int): The alpha value for the RGBA color (0-255).
|
73 |
-
|
74 |
-
Returns:
|
75 |
-
tuple: A tuple representing the RGBA color values (red, green, blue, alpha).
|
76 |
-
"""
|
77 |
-
hex_color = hex_color.lstrip("#")
|
78 |
-
r, g, b = int(hex_color[0:2], 16), int(hex_color[2:4], 16), int(hex_color[4:6], 16)
|
79 |
-
return (r, g, b, alpha)
|
80 |
-
|
81 |
-
|
82 |
-
# Draw OCR bounding boxes with enhanced visual elements
|
83 |
-
def draw_ocr_bboxes(image, prediction):
|
84 |
-
"""
|
85 |
-
Draw bounding boxes with enhanced visual elements on the given image based on the OCR prediction.
|
86 |
-
|
87 |
-
Args:
|
88 |
-
image (PIL.Image.Image): The input image on which the bounding boxes will be drawn.
|
89 |
-
prediction (dict): The OCR prediction containing 'quad_boxes' and 'labels'.
|
90 |
-
|
91 |
-
Returns:
|
92 |
-
PIL.Image.Image: The image with the bounding boxes drawn.
|
93 |
-
"""
|
94 |
-
|
95 |
-
# Create a drawing object for the image with RGBA mode
|
96 |
-
draw = ImageDraw.Draw(image, "RGBA")
|
97 |
-
|
98 |
-
# Extract bounding boxes and labels from the prediction
|
99 |
-
bboxes, labels = prediction["quad_boxes"], prediction["labels"]
|
100 |
-
|
101 |
-
for i, (box, label) in enumerate(zip(bboxes, labels)):
|
102 |
-
# Select color for the bounding box and label
|
103 |
-
color = colormap[i % len(colormap)]
|
104 |
-
new_box = (np.array(box)).tolist()
|
105 |
-
|
106 |
-
# Define the outline width and corner radius for the bounding box
|
107 |
-
box_outline_width = 3
|
108 |
-
corner_radius = 10
|
109 |
-
|
110 |
-
# Draw rounded corners for the bounding box
|
111 |
-
for j in range(4):
|
112 |
-
start_x, start_y = new_box[j * 2], new_box[j * 2 + 1]
|
113 |
-
end_x, end_y = new_box[(j * 2 + 2) % 8], new_box[(j * 2 + 3) % 8]
|
114 |
-
|
115 |
-
# Draw the arcs for the rounded corners
|
116 |
-
draw.arc(
|
117 |
-
[
|
118 |
-
(start_x - corner_radius, start_y - corner_radius),
|
119 |
-
(start_x + corner_radius, start_y + corner_radius),
|
120 |
-
],
|
121 |
-
90 + j * 90,
|
122 |
-
180 + j * 90,
|
123 |
-
fill=color,
|
124 |
-
width=box_outline_width,
|
125 |
-
)
|
126 |
-
draw.arc(
|
127 |
-
[
|
128 |
-
(end_x - corner_radius, end_y - corner_radius),
|
129 |
-
(end_x + corner_radius, end_y + corner_radius),
|
130 |
-
],
|
131 |
-
j * 90,
|
132 |
-
90 + j * 90,
|
133 |
-
fill=color,
|
134 |
-
width=box_outline_width,
|
135 |
-
)
|
136 |
-
|
137 |
-
# Draw the lines connecting the arcs
|
138 |
-
if j in [0, 1, 2]:
|
139 |
-
draw.line(
|
140 |
-
[
|
141 |
-
(start_x + corner_radius if j != 1 else start_x, start_y),
|
142 |
-
(end_x - corner_radius if j != 1 else end_x, end_y),
|
143 |
-
],
|
144 |
-
fill=color,
|
145 |
-
width=box_outline_width,
|
146 |
-
)
|
147 |
-
else:
|
148 |
-
draw.line(
|
149 |
-
[
|
150 |
-
(start_x, start_y + corner_radius),
|
151 |
-
(end_x, end_y - corner_radius),
|
152 |
-
],
|
153 |
-
fill=color,
|
154 |
-
width=box_outline_width,
|
155 |
-
)
|
156 |
-
|
157 |
-
# Calculate the position for the text label
|
158 |
-
text_x, text_y = min(new_box[0::2]), min(new_box[1::2]) - 20
|
159 |
-
text_w, text_h = draw.
|
160 |
-
rgba_color = hex_to_rgba(color, 200) # Semi-transparent background for text
|
161 |
-
|
162 |
-
# Draw the background rectangle for the text
|
163 |
-
draw.rectangle(
|
164 |
-
[text_x, text_y, text_x + text_w + 10, text_y + text_h + 10],
|
165 |
-
fill=rgba_color,
|
166 |
-
)
|
167 |
-
|
168 |
-
# Draw the text label
|
169 |
-
draw.text((text_x + 5, text_y + 5), label, fill=(0, 0, 0, 255))
|
170 |
-
|
171 |
-
# Return the image with the OCR boxes drawn
|
172 |
-
return image
|
|
|
1 |
+
from PIL import ImageDraw
|
2 |
+
import numpy as np
|
3 |
+
import re
|
4 |
+
|
5 |
+
|
6 |
+
# Use a color map for bounding boxes
|
7 |
+
colormap = [
|
8 |
+
"#0000FF",
|
9 |
+
"#FFA500",
|
10 |
+
"#008000",
|
11 |
+
"#800080",
|
12 |
+
"#A52A2A",
|
13 |
+
"#FFC0CB",
|
14 |
+
"#808080",
|
15 |
+
"#808000",
|
16 |
+
"#00FFFF",
|
17 |
+
"#FF0000",
|
18 |
+
"#00FF00",
|
19 |
+
"#4B0082",
|
20 |
+
"#4B0082",
|
21 |
+
"#EE82EE",
|
22 |
+
"#00FFFF",
|
23 |
+
"#FF00FF",
|
24 |
+
"#FF7F50",
|
25 |
+
"#FFD700",
|
26 |
+
"#87CEEB",
|
27 |
+
]
|
28 |
+
|
29 |
+
|
30 |
+
# Text cleaning function
|
31 |
+
def clean_text(text):
|
32 |
+
"""
|
33 |
+
Cleans the given text by removing unwanted tokens, extra spaces,
|
34 |
+
and ensures proper spacing between words and after periods.
|
35 |
+
|
36 |
+
Args:
|
37 |
+
text (str): The input text to be cleaned.
|
38 |
+
|
39 |
+
Returns:
|
40 |
+
str: The cleaned and properly formatted text.
|
41 |
+
"""
|
42 |
+
|
43 |
+
# Remove unwanted tokens
|
44 |
+
text = text.replace("<pad>", "").replace("</s>", "").strip()
|
45 |
+
|
46 |
+
# Split the text into lines and clean each line
|
47 |
+
lines = text.split("\n")
|
48 |
+
cleaned_lines = [line.strip() for line in lines if line.strip()]
|
49 |
+
|
50 |
+
# Join the cleaned lines into a single string with a space between each line
|
51 |
+
cleaned_text = " ".join(cleaned_lines)
|
52 |
+
|
53 |
+
# Ensure proper spacing between words and after periods using regex
|
54 |
+
cleaned_text = re.sub(
|
55 |
+
r"\s+", " ", cleaned_text
|
56 |
+
) # Replace multiple spaces with a single space
|
57 |
+
cleaned_text = re.sub(
|
58 |
+
r"(?<=[.])(?=[^\s])", r" ", cleaned_text
|
59 |
+
) # Add space after a period if not followed by a space
|
60 |
+
|
61 |
+
# Return the cleaned text
|
62 |
+
return cleaned_text
|
63 |
+
|
64 |
+
|
65 |
+
# Convert hex color to RGBA with the given alpha
|
66 |
+
def hex_to_rgba(hex_color, alpha):
|
67 |
+
"""
|
68 |
+
Convert a hexadecimal color code to RGBA format.
|
69 |
+
|
70 |
+
Args:
|
71 |
+
hex_color (str): The hexadecimal color code (e.g., "#FF0000").
|
72 |
+
alpha (int): The alpha value for the RGBA color (0-255).
|
73 |
+
|
74 |
+
Returns:
|
75 |
+
tuple: A tuple representing the RGBA color values (red, green, blue, alpha).
|
76 |
+
"""
|
77 |
+
hex_color = hex_color.lstrip("#")
|
78 |
+
r, g, b = int(hex_color[0:2], 16), int(hex_color[2:4], 16), int(hex_color[4:6], 16)
|
79 |
+
return (r, g, b, alpha)
|
80 |
+
|
81 |
+
|
82 |
+
# Draw OCR bounding boxes with enhanced visual elements
|
83 |
+
def draw_ocr_bboxes(image, prediction):
|
84 |
+
"""
|
85 |
+
Draw bounding boxes with enhanced visual elements on the given image based on the OCR prediction.
|
86 |
+
|
87 |
+
Args:
|
88 |
+
image (PIL.Image.Image): The input image on which the bounding boxes will be drawn.
|
89 |
+
prediction (dict): The OCR prediction containing 'quad_boxes' and 'labels'.
|
90 |
+
|
91 |
+
Returns:
|
92 |
+
PIL.Image.Image: The image with the bounding boxes drawn.
|
93 |
+
"""
|
94 |
+
|
95 |
+
# Create a drawing object for the image with RGBA mode
|
96 |
+
draw = ImageDraw.Draw(image, "RGBA")
|
97 |
+
|
98 |
+
# Extract bounding boxes and labels from the prediction
|
99 |
+
bboxes, labels = prediction["quad_boxes"], prediction["labels"]
|
100 |
+
|
101 |
+
for i, (box, label) in enumerate(zip(bboxes, labels)):
|
102 |
+
# Select color for the bounding box and label
|
103 |
+
color = colormap[i % len(colormap)]
|
104 |
+
new_box = (np.array(box)).tolist()
|
105 |
+
|
106 |
+
# Define the outline width and corner radius for the bounding box
|
107 |
+
box_outline_width = 3
|
108 |
+
corner_radius = 10
|
109 |
+
|
110 |
+
# Draw rounded corners for the bounding box
|
111 |
+
for j in range(4):
|
112 |
+
start_x, start_y = new_box[j * 2], new_box[j * 2 + 1]
|
113 |
+
end_x, end_y = new_box[(j * 2 + 2) % 8], new_box[(j * 2 + 3) % 8]
|
114 |
+
|
115 |
+
# Draw the arcs for the rounded corners
|
116 |
+
draw.arc(
|
117 |
+
[
|
118 |
+
(start_x - corner_radius, start_y - corner_radius),
|
119 |
+
(start_x + corner_radius, start_y + corner_radius),
|
120 |
+
],
|
121 |
+
90 + j * 90,
|
122 |
+
180 + j * 90,
|
123 |
+
fill=color,
|
124 |
+
width=box_outline_width,
|
125 |
+
)
|
126 |
+
draw.arc(
|
127 |
+
[
|
128 |
+
(end_x - corner_radius, end_y - corner_radius),
|
129 |
+
(end_x + corner_radius, end_y + corner_radius),
|
130 |
+
],
|
131 |
+
j * 90,
|
132 |
+
90 + j * 90,
|
133 |
+
fill=color,
|
134 |
+
width=box_outline_width,
|
135 |
+
)
|
136 |
+
|
137 |
+
# Draw the lines connecting the arcs
|
138 |
+
if j in [0, 1, 2]:
|
139 |
+
draw.line(
|
140 |
+
[
|
141 |
+
(start_x + corner_radius if j != 1 else start_x, start_y),
|
142 |
+
(end_x - corner_radius if j != 1 else end_x, end_y),
|
143 |
+
],
|
144 |
+
fill=color,
|
145 |
+
width=box_outline_width,
|
146 |
+
)
|
147 |
+
else:
|
148 |
+
draw.line(
|
149 |
+
[
|
150 |
+
(start_x, start_y + corner_radius),
|
151 |
+
(end_x, end_y - corner_radius),
|
152 |
+
],
|
153 |
+
fill=color,
|
154 |
+
width=box_outline_width,
|
155 |
+
)
|
156 |
+
|
157 |
+
# Calculate the position for the text label
|
158 |
+
text_x, text_y = min(new_box[0::2]), min(new_box[1::2]) - 20
|
159 |
+
text_w, text_h = draw.textlength(label)
|
160 |
+
rgba_color = hex_to_rgba(color, 200) # Semi-transparent background for text
|
161 |
+
|
162 |
+
# Draw the background rectangle for the text
|
163 |
+
draw.rectangle(
|
164 |
+
[text_x, text_y, text_x + text_w + 10, text_y + text_h + 10],
|
165 |
+
fill=rgba_color,
|
166 |
+
)
|
167 |
+
|
168 |
+
# Draw the text label
|
169 |
+
draw.text((text_x + 5, text_y + 5), label, fill=(0, 0, 0, 255))
|
170 |
+
|
171 |
+
# Return the image with the OCR boxes drawn
|
172 |
+
return image
|