-
Notifications
You must be signed in to change notification settings - Fork 15
/
template_model.py
56 lines (48 loc) · 2.59 KB
/
template_model.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
"""Model class template
This module provides a template for users to implement custom models.
The filename should be <model>_model.py
The class name should be <Model>Model
It implements a simple multi-mapping translation baseline with a unified model.
You need to implement the following functions:
<prepare_data>: Unpack input data and perform pre-processing steps.
<translation>: Perform image translation for model evalution.
"""
import torch
from models.base_model import BaseModel
################## TemplateModel #############################
class TemplateModel(BaseModel):
def __init__(self, opt):
BaseModel.__init__(self, opt)
def prepare_data(self, data):
'''prepare data for training or inference.
Parameters:
data -- It should contain the image and the corresponding domain information.
Returns:
img -- Input image.
domain_source -- Domain information of the input image, such as one-hot lable, attribute lable, or sentence embedding.
index_target -- Define the target mapping of the input image.
weight_source -- Define the domain weight to tackle the domain(class) imbalance problem.
'''
img, domain_source = data
index_target = torch.randperm(img.size(0))
weight_source = torch.ones([img.size(0),1])
self.current_data = [img.to(self.device),
domain_source.to(self.device),
index_target.to(self.device),
weight_source.to(self.device)]
return self.current_data
def translation(self, data):
'''translate the input image'''
with torch.no_grad():
img, domain_source, index_target = self.prepare_data(data)
domain_target = domain_source[index_target]
style_enc, _, _ = self.enc_style(img)
style_rand = self.sample_latent_code(style_enc.size())
content = self.enc_content(img)
enc_rec = self.dec(content,torch.cat([domain_source,style_enc],dim=1))
rand_intra_domain = self.dec(content,torch.cat([domain_source,style_rand],dim=1))
rand_inter_domain = self.dec(content,torch.cat([domain_target,style_rand],dim=1))
return [('input': tensor2im(img.data)),
('reconstruction': tensor2im(enc_rec.data)),
('intra-domain translation': tensor2im(rand_intra_domain.data)),
('inter-domain translation': tensor2im(rand_inter_domain.data))]