tejasexpress
commited on
utility scripts
Browse files- dataframe.py +61 -0
- json-to-annotation.py +39 -0
dataframe.py
ADDED
@@ -0,0 +1,61 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import os
|
2 |
+
import pandas as pd
|
3 |
+
from PIL import Image
|
4 |
+
from io import BytesIO
|
5 |
+
import io
|
6 |
+
|
7 |
+
|
8 |
+
# Path to the folders containing images and captions
|
9 |
+
image_folder = "images"
|
10 |
+
ocr_folder = "ocr-text"
|
11 |
+
|
12 |
+
# List to store data
|
13 |
+
data = []
|
14 |
+
count = 0
|
15 |
+
|
16 |
+
# Iterate through images and captions
|
17 |
+
for index, image_file in enumerate(os.listdir(image_folder)):
|
18 |
+
if index >= 1000:
|
19 |
+
break
|
20 |
+
count += 1
|
21 |
+
print("adding image: " + image_file + " to dataframe" + " count: " + str(count))
|
22 |
+
image_path = os.path.join(image_folder, image_file)
|
23 |
+
|
24 |
+
# Assuming caption file names match image file names
|
25 |
+
ocr_path = os.path.join(ocr_folder, image_file.replace(".jpg", "_words.txt"))
|
26 |
+
|
27 |
+
# Read image and get image dimensions
|
28 |
+
image = Image.open(image_path)
|
29 |
+
image_width, image_height = image.size
|
30 |
+
|
31 |
+
# Read caption
|
32 |
+
with open(ocr_path, "r", encoding="utf-8") as ocr_file:
|
33 |
+
caption = ocr_file.read().strip()
|
34 |
+
|
35 |
+
# Convert image to byte array
|
36 |
+
img = Image.open(image_path)
|
37 |
+
img_byte_arr = io.BytesIO()
|
38 |
+
img.save(img_byte_arr, format=img.format)
|
39 |
+
img_byte_arr = img_byte_arr.getvalue()
|
40 |
+
|
41 |
+
# Append data to the list
|
42 |
+
data.append({
|
43 |
+
'image': img_byte_arr,
|
44 |
+
'ocr_annotation_texts': caption,
|
45 |
+
'image_height': image_height,
|
46 |
+
'image_width': image_width
|
47 |
+
})
|
48 |
+
|
49 |
+
# Create DataFrame
|
50 |
+
df = pd.DataFrame(data)
|
51 |
+
|
52 |
+
current_directory = os.getcwd()
|
53 |
+
|
54 |
+
parquet_file_name = "my_dataframe.parquet"
|
55 |
+
|
56 |
+
# Combine the current directory and the filename to create the full path
|
57 |
+
parquet_file_path = os.path.join(current_directory, parquet_file_name)
|
58 |
+
df.to_parquet(parquet_file_path, index=False)
|
59 |
+
|
60 |
+
# Display the DataFrame
|
61 |
+
print(df)
|
json-to-annotation.py
ADDED
@@ -0,0 +1,39 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import os
|
2 |
+
import json
|
3 |
+
|
4 |
+
input_folder = "words"
|
5 |
+
output_folder = "ocr-text"
|
6 |
+
|
7 |
+
json_files = [file for file in os.listdir(input_folder) if file.endswith('.json')]
|
8 |
+
|
9 |
+
for json_file in json_files:
|
10 |
+
print(f'Processing {json_file}')
|
11 |
+
input_file_path = os.path.join(input_folder, json_file)
|
12 |
+
output_file_path = os.path.join(output_folder, os.path.splitext(json_file)[0] + '.txt')
|
13 |
+
|
14 |
+
with open(input_file_path, 'r') as file:
|
15 |
+
data = json.load(file)
|
16 |
+
|
17 |
+
words_list = data.get("words", [])
|
18 |
+
|
19 |
+
#calculate width and height of the image
|
20 |
+
width = data["image_rect"][2]
|
21 |
+
height = data["image_rect"][3]
|
22 |
+
|
23 |
+
result_strings = []
|
24 |
+
for word in words_list:
|
25 |
+
bbox_values = word["bbox"]
|
26 |
+
|
27 |
+
#calculate (x,y,w,h)
|
28 |
+
x = int((bbox_values[0])*100/width)
|
29 |
+
y = int((bbox_values[1])*100/height)
|
30 |
+
box_width = int((abs(bbox_values[2]-bbox_values[0]))*100/width)
|
31 |
+
box_height = int((abs(bbox_values[3]-bbox_values[1]))*100/height)
|
32 |
+
#convert everything to a string with appropriate spacing and text at the end
|
33 |
+
result_string = f"{x} {y} {box_width} {box_height} {word['text']}"
|
34 |
+
result_strings.append(result_string)
|
35 |
+
|
36 |
+
result_string = ' '.join(result_strings)
|
37 |
+
|
38 |
+
with open(output_file_path, 'w', encoding='utf-8') as file:
|
39 |
+
file.write(result_string)
|