-
Notifications
You must be signed in to change notification settings - Fork 887
/
detect_layout.py
73 lines (58 loc) · 3.23 KB
/
detect_layout.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
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
import time
import pypdfium2 # Causes a warning if not the top import
import argparse
import copy
import json
from collections import defaultdict
from surya.detection import batch_text_detection
from surya.input.load import load_from_folder, load_from_file
from surya.layout import batch_layout_detection
from surya.model.detection.model import load_model as load_det_model, load_processor as load_det_processor
from surya.model.layout.model import load_model, load_processor
from surya.postprocessing.heatmap import draw_polys_on_image
from surya.settings import settings
import os
def main():
parser = argparse.ArgumentParser(description="Detect layout of an input file or folder (PDFs or image).")
parser.add_argument("input_path", type=str, help="Path to pdf or image file or folder to detect layout in.")
parser.add_argument("--results_dir", type=str, help="Path to JSON file with layout results.", default=os.path.join(settings.RESULT_DIR, "surya"))
parser.add_argument("--max", type=int, help="Maximum number of pages to process.", default=None)
parser.add_argument("--images", action="store_true", help="Save images of detected layout bboxes.", default=False)
parser.add_argument("--debug", action="store_true", help="Run in debug mode.", default=False)
args = parser.parse_args()
model = load_model()
processor = load_processor()
det_model = load_det_model()
det_processor = load_det_processor()
if os.path.isdir(args.input_path):
images, names, _ = load_from_folder(args.input_path, args.max)
folder_name = os.path.basename(args.input_path)
else:
images, names, _ = load_from_file(args.input_path, args.max)
folder_name = os.path.basename(args.input_path).split(".")[0]
start = time.time()
line_predictions = batch_text_detection(images, det_model, det_processor)
layout_predictions = batch_layout_detection(images, model, processor, line_predictions, include_maps=args.debug)
result_path = os.path.join(args.results_dir, folder_name)
os.makedirs(result_path, exist_ok=True)
if args.debug:
print(f"Layout took {time.time() - start} seconds")
if args.images:
for idx, (image, layout_pred, name) in enumerate(zip(images, layout_predictions, names)):
polygons = [p.polygon for p in layout_pred.bboxes]
labels = [p.label for p in layout_pred.bboxes]
bbox_image = draw_polys_on_image(polygons, copy.deepcopy(image), labels=labels)
bbox_image.save(os.path.join(result_path, f"{name}_{idx}_layout.png"))
if args.debug:
heatmap = layout_pred.segmentation_map
heatmap.save(os.path.join(result_path, f"{name}_{idx}_segmentation.png"))
predictions_by_page = defaultdict(list)
for idx, (pred, name, image) in enumerate(zip(layout_predictions, names, images)):
out_pred = pred.model_dump(exclude=["segmentation_map"])
out_pred["page"] = len(predictions_by_page[name]) + 1
predictions_by_page[name].append(out_pred)
with open(os.path.join(result_path, "results.json"), "w+", encoding="utf-8") as f:
json.dump(predictions_by_page, f, ensure_ascii=False)
print(f"Wrote results to {result_path}")
if __name__ == "__main__":
main()