Upload sentence-annotation.py
Browse files- sentence-annotation.py +71 -0
sentence-annotation.py
ADDED
|
@@ -0,0 +1,71 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
import os
|
| 2 |
+
import json
|
| 3 |
+
import matplotlib.pyplot as plt
|
| 4 |
+
import matplotlib.patches as patches
|
| 5 |
+
|
| 6 |
+
|
| 7 |
+
input_folder = "words"
|
| 8 |
+
output_folder = "ocr-text-sentence"
|
| 9 |
+
|
| 10 |
+
json_files = [file for file in os.listdir(input_folder) if file.endswith('.json')]
|
| 11 |
+
|
| 12 |
+
for idx, json_file in enumerate(json_files):
|
| 13 |
+
print(f'Processing {json_file}')
|
| 14 |
+
input_file_path = os.path.join(input_folder, json_file)
|
| 15 |
+
output_file_path = os.path.join(output_folder, os.path.splitext(json_file)[0] + '.txt')
|
| 16 |
+
img_file_path = os.path.join("images", json_file.replace("_words.json", ".jpg"))
|
| 17 |
+
|
| 18 |
+
with open(input_file_path, 'r') as file:
|
| 19 |
+
data = json.load(file)
|
| 20 |
+
|
| 21 |
+
words_list = data.get("words", [])
|
| 22 |
+
|
| 23 |
+
width = data["image_rect"][2]
|
| 24 |
+
height = data["image_rect"][3]
|
| 25 |
+
print(len(words_list))
|
| 26 |
+
sorted_list = sorted(words_list, key=lambda x: (x["bbox"][1], x["bbox"][0]))
|
| 27 |
+
print(sum(len(s["text"]) for s in sorted_list))
|
| 28 |
+
|
| 29 |
+
merged_list = []
|
| 30 |
+
for i, item in enumerate(sorted_list):
|
| 31 |
+
if i > 0:
|
| 32 |
+
prev_item = sorted_list[i - 1]
|
| 33 |
+
# Check if the current item is in the same line as the previous one
|
| 34 |
+
if abs(item["bbox"][1] - prev_item["bbox"][1]) < 4 and (item["bbox"][0] - prev_item["bbox"][2])<10:
|
| 35 |
+
# Merge the boxes and update the text
|
| 36 |
+
merged_item = {
|
| 37 |
+
"bbox": [min(item["bbox"][0], merged_list[-1]["bbox"][0]),
|
| 38 |
+
min(item["bbox"][1], merged_list[-1]["bbox"][1]),
|
| 39 |
+
max(item["bbox"][2], merged_list[-1]["bbox"][2]),
|
| 40 |
+
max(item["bbox"][3], merged_list[-1]["bbox"][3])],
|
| 41 |
+
"text": merged_list[-1]["text"] + " " + item["text"]
|
| 42 |
+
}
|
| 43 |
+
# Update the previous item in the merged list
|
| 44 |
+
merged_list[-1] = merged_item
|
| 45 |
+
else:
|
| 46 |
+
# Add the current item to the merged list
|
| 47 |
+
merged_list.append(item)
|
| 48 |
+
else:
|
| 49 |
+
# Add the first item to the merged list
|
| 50 |
+
merged_list.append(item)
|
| 51 |
+
|
| 52 |
+
print(sum(len(s["text"]) for s in merged_list))
|
| 53 |
+
result_strings = []
|
| 54 |
+
for word in merged_list:
|
| 55 |
+
bbox_values = word["bbox"]
|
| 56 |
+
|
| 57 |
+
#calculate (x,y,w,h)
|
| 58 |
+
x = int((bbox_values[0])*100/width)
|
| 59 |
+
y = int((bbox_values[1])*100/height)
|
| 60 |
+
box_width = int((abs(bbox_values[2]-bbox_values[0]))*100/width)
|
| 61 |
+
box_height = int((abs(bbox_values[3]-bbox_values[1]))*100/height)
|
| 62 |
+
#convert everything to a string with appropriate spacing and text at the end
|
| 63 |
+
result_string = f"{x} {y} {box_width} {box_height} {word['text']}"
|
| 64 |
+
result_strings.append(result_string)
|
| 65 |
+
|
| 66 |
+
result_string = ' '.join(result_strings)
|
| 67 |
+
|
| 68 |
+
with open(output_file_path, 'w', encoding='utf-8') as file:
|
| 69 |
+
file.write(result_string)
|
| 70 |
+
|
| 71 |
+
|