-
Notifications
You must be signed in to change notification settings - Fork 47
/
format_docsynth300k.py
37 lines (33 loc) · 1.33 KB
/
format_docsynth300k.py
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
import io
import os
import tqdm
import pandas as pd
from PIL import Image
parquet_list = os.listdir("./docsynth300k")
parquet_list = [f for f in parquet_list if "parquet" in f]
save_path = "./layout_data/docsynth300k"
if not os.path.exists(save_path):
os.makedirs(save_path)
os.makedirs(os.path.join(save_path, "images"))
os.makedirs(os.path.join(save_path, "labels"))
train_txt = open(os.path.join(save_path, "train300k.txt"), "w")
for parquet in tqdm.tqdm(parquet_list):
df = pd.read_parquet(os.path.join("docsynth300k", parquet))
for index, row in tqdm.tqdm(df.iterrows()):
filename = row['filename']
image_data = row['image_data']
anno_string = row['anno_string']
image = Image.open(io.BytesIO(image_data))
# save image / anno / txt
train_txt.write(os.path.join("images", filename) + "\n")
image.save(os.path.join(save_path, "images", filename))
with open(os.path.join(save_path, "labels", filename.replace(".jpg", ".txt")), "w") as f:
for line in anno_string:
f.write(line)
# write validation file (dummy)
train_txt.close()
image_list = list(open(os.path.join(save_path, "train300k.txt"), "r").readlines())[:1000]
val = open(os.path.join(save_path, "val.txt"), "w")
for line in image_list:
val.write(line)
val.close()