parralel processing
Browse files- collect-dataset.py +59 -28
collect-dataset.py
CHANGED
@@ -6,7 +6,10 @@ from typing import Dict, List
|
|
6 |
from io import BytesIO
|
7 |
import requests
|
8 |
from PIL import Image
|
|
|
9 |
from pytesseract import pytesseract
|
|
|
|
|
10 |
|
11 |
# im = Image.MAGICK
|
12 |
# im.magick_init()
|
@@ -21,6 +24,11 @@ def fetch_image(url):
|
|
21 |
image_content = response.content
|
22 |
return Image.open(BytesIO(image_content))
|
23 |
|
|
|
|
|
|
|
|
|
|
|
24 |
|
25 |
# def modify_image(
|
26 |
# image_buffer: bytes,
|
@@ -77,37 +85,60 @@ def get_orgs(min_rank: int) -> List[Dict[str, any]]:
|
|
77 |
} for entity in res_json.get("entities", [])
|
78 |
]
|
79 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
80 |
|
81 |
-
|
82 |
shutil.rmtree("./data/", ignore_errors=True)
|
83 |
os.mkdir("./data/")
|
84 |
-
counter = 0
|
85 |
start_page = int(sys.argv[1])
|
|
|
86 |
for i in range(start_page, 3000):
|
87 |
print(f"Processing page {i}...")
|
88 |
-
|
89 |
-
|
90 |
-
|
91 |
-
|
92 |
-
|
93 |
-
|
94 |
-
|
95 |
-
|
96 |
-
|
97 |
-
|
98 |
-
|
99 |
-
|
100 |
-
|
101 |
-
|
102 |
-
|
103 |
-
|
104 |
-
|
105 |
-
|
106 |
-
|
107 |
-
|
108 |
-
org.get("name"), org.get("short_description").replace("{} is ".format(org.get("name")), ""))
|
109 |
-
print("{} - {}".format(counter, org.get("name")))
|
110 |
-
with open("./data/metadata.jsonl", "a") as f:
|
111 |
-
f.write(json.dumps(
|
112 |
-
{"file_name": f"{org.get('uuid')}.png", "text": text}) + "\n")
|
113 |
-
counter += 1
|
|
|
6 |
from io import BytesIO
|
7 |
import requests
|
8 |
from PIL import Image
|
9 |
+
import asyncio
|
10 |
from pytesseract import pytesseract
|
11 |
+
import concurrent.futures
|
12 |
+
|
13 |
|
14 |
# im = Image.MAGICK
|
15 |
# im.magick_init()
|
|
|
24 |
image_content = response.content
|
25 |
return Image.open(BytesIO(image_content))
|
26 |
|
27 |
+
def convert_alpha_to_white(img):
|
28 |
+
if img.mode != 'RGBA':
|
29 |
+
return img
|
30 |
+
white_bg = Image.new('RGBA', img.size, (255, 255, 255, 255))
|
31 |
+
return Image.alpha_composite(white_bg, img)
|
32 |
|
33 |
# def modify_image(
|
34 |
# image_buffer: bytes,
|
|
|
85 |
} for entity in res_json.get("entities", [])
|
86 |
]
|
87 |
|
88 |
+
def handle_org(org):
|
89 |
+
print(f"Processing {org.get('name')}...")
|
90 |
+
try:
|
91 |
+
image = fetch_image(
|
92 |
+
org.get("image_url"))
|
93 |
+
if (image.format != "PNG" and image.format != "JPEG"):
|
94 |
+
raise ValueError("Image is not PNG or JPEG")
|
95 |
+
width, height = image.size
|
96 |
+
if width != height:
|
97 |
+
raise ValueError("Image is not square")
|
98 |
+
if (width < 400 or height < 400):
|
99 |
+
raise ValueError(f"Image is too small - {width}x{height}")
|
100 |
+
image = convert_alpha_to_white(image)
|
101 |
+
image = image.resize((512, 512), resample=Image.LANCZOS)
|
102 |
+
image_text = pytesseract.image_to_string(image, config='--psm 11')
|
103 |
+
if (len(image_text) >= 3):
|
104 |
+
raise ValueError("Image has text")
|
105 |
+
image.save(f"./data/{org.get('uuid')}.png", format="PNG")
|
106 |
+
except Exception as ex:
|
107 |
+
print(
|
108 |
+
f"Error downloading image for {org.get('uuid')}: {ex}")
|
109 |
+
return
|
110 |
+
text = "Logo of {}, {}".format(
|
111 |
+
org.get("name"), org.get("short_description").replace("{} is ".format(org.get("name")), ""))
|
112 |
+
print("{}".format( org.get("name")))
|
113 |
+
with open("./data/metadata.jsonl", "a") as f:
|
114 |
+
f.write(json.dumps(
|
115 |
+
{"file_name": f"{org.get('uuid')}.png", "text": text}) + "\n")
|
116 |
+
return {"file_name": f"{org.get('uuid')}.png", "text": text}
|
117 |
|
118 |
+
def main():
|
119 |
shutil.rmtree("./data/", ignore_errors=True)
|
120 |
os.mkdir("./data/")
|
|
|
121 |
start_page = int(sys.argv[1])
|
122 |
+
|
123 |
for i in range(start_page, 3000):
|
124 |
print(f"Processing page {i}...")
|
125 |
+
with concurrent.futures.ThreadPoolExecutor() as executor:
|
126 |
+
futures = []
|
127 |
+
orgs = get_orgs(i*1000)
|
128 |
+
print(f"Found {len(orgs)} orgs")
|
129 |
+
for org in orgs:
|
130 |
+
futures.append(executor.submit(handle_org, org))
|
131 |
+
# futures.append(executor.submit(handle_org, org) for org in orgs)
|
132 |
+
for future in concurrent.futures.as_completed(futures):
|
133 |
+
result = future.result()
|
134 |
+
if result is not None:
|
135 |
+
print(result)
|
136 |
+
if (len(orgs) < 1001):
|
137 |
+
print("No more orgs")
|
138 |
+
break
|
139 |
+
|
140 |
+
|
141 |
+
|
142 |
+
|
143 |
+
if __name__ == "__main__":
|
144 |
+
main()
|
|
|
|
|
|
|
|
|
|
|
|