chulanpro5 commited on
Commit
79378ee
·
1 Parent(s): 1edeb9d

update: add simple inference example

Browse files
Files changed (1) hide show
  1. simple_inference.py +76 -0
simple_inference.py ADDED
@@ -0,0 +1,76 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ import random
2
+ from typing import List, Union, Optional, Tuple
3
+ import torch
4
+ from PIL import Image
5
+ from sample import (arg_parse,
6
+ sampling,
7
+ load_fontdiffuer_pipeline)
8
+
9
+ def run_fontdiffuer(source_image,
10
+ character,
11
+ reference_image,
12
+ sampling_step,
13
+ guidance_scale,
14
+ batch_size=1):
15
+
16
+ args.character_input = False if source_image is not None else True
17
+ args.content_character = character
18
+ args.sampling_step = sampling_step
19
+ args.guidance_scale = guidance_scale
20
+ args.batch_size = batch_size
21
+ args.seed = random.randint(0, 10000)
22
+ out_image = sampling(
23
+ args=args,
24
+ pipe=pipe,
25
+ content_image=source_image,
26
+ style_image=reference_image)
27
+
28
+ if out_image is not None:
29
+ out_image.format = 'PNG'
30
+
31
+ return out_image
32
+
33
+ def run_inference(
34
+ source_image_path: Union[str, None],
35
+ character: Union[str, None],
36
+ reference_image_path: str,
37
+ sampling_step: int=50,
38
+ guidance_scale: float=7.5,
39
+ ):
40
+ if source_image_path is not None:
41
+ source_image = Image.open(source_image_path).convert('RGB')
42
+ else:
43
+ source_image = None
44
+
45
+ if reference_image_path is not None:
46
+ reference_image = Image.open(reference_image_path).convert('RGB')
47
+ else:
48
+ reference_image = None
49
+
50
+ image = run_fontdiffuer(
51
+ source_image=source_image,
52
+ character=character,
53
+ reference_image=reference_image,
54
+ sampling_step=sampling_step,
55
+ guidance_scale=guidance_scale
56
+ )
57
+ return image
58
+
59
+ if __name__ == '__main__':
60
+ args = arg_parse()
61
+ args.demo = True
62
+ args.ckpt_dir = 'ckpt'
63
+ args.ttf_path = 'ttf/KaiXinSongA.ttf'
64
+ args.device = 'cuda'
65
+
66
+
67
+ # load fontdiffuer pipeline
68
+ pipe = load_fontdiffuer_pipeline(args=args)
69
+
70
+ image = run_inference(
71
+ character=None,
72
+ source_image_path="figures/ref_imgs/ref_壤.jpg",
73
+ reference_image_path="figures/ref_imgs/ref_欟.jpg"
74
+ )
75
+
76
+ print(image)