-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathapply-demosaic.py
148 lines (121 loc) · 5.36 KB
/
apply-demosaic.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
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
import argparse
import os
import torch
import numpy as np
from data import ProDemosaicDataset, SharpDemosaicDataset, N2SDataset
import matplotlib.pyplot as plt
from PIL import Image
import json
import model
if __name__=="__main__":
parser = argparse.ArgumentParser()
parser.add_argument("model")
parser.add_argument("inputdir")
parser.add_argument("outdir")
parser.add_argument("--config", type=str, nargs="?", default="", help="Use the configs the model was trained with to load it")
parser.add_argument("--class", type=str, nargs="?", default="unet", help="which model class to use")
parser.add_argument("--statedict", type=str, nargs="?", const=True, default=False, help="whether the model save is a state dict. Assumes pickle elsewise.")
parser.add_argument("--noconvert", nargs="?", type=bool, default=False, const=True)
parser.add_argument("--dataset", type=str, default="pro", help="Type of dataset, sharp or pro")
parser.add_argument("--device", type=str, default="cpu", help="The device to use for inference")
parser.add_argument("--channelswap", type=bool, nargs="?", default=False, const=True, help="Use for older models that were trained on the dataset that swaps channels")
parser.add_argument("--channels", type=int, default=2, help="number of denoising channels" )
parser.add_argument("--save_input", action="store_true")
parser.add_argument("--save_gt", action="store_true")
parser.add_argument("--chop", type=int, default=1)
parser.add_argument("--n2s", action="store_true")
args = parser.parse_args()
channels = args.channels
if not os.path.exists(args.inputdir):
raise ValueError("Input directory not found: {}".format(args.inputdir))
if not os.path.exists(args.outdir):
os.mkdir(args.outdir)
config = {}
if not args.config == "":
with open(args.config, 'r') as f:
config = json.load(f)
if args.dataset == "pro":
dataset = ProDemosaicDataset(
args.inputdir,
crop=False,
patches_per_image=1)
elif args.dataset == "sharp":
dataset = SharpDemosaicDataset(
args.inputdir,
crop=False)
elif args.dataset == "direct_sharp":
dataset = N2SDataset(
args.inputdir,
crop=False,
sharp=True,
channels=channels,
patches_per_image=1)
state = torch.load(args.model, map_location=args.device)
if args.statedict:
model_name = config.get("model", "unet")
model_params = config.get("model_params", {})
if model_name == "resnet":
net = model.ResNet(2, 2, **model_params)
elif model_name == "unet":
net = model.UNet(2, **model_params)
elif model_name == "n2s-unet":
from noise2self.models.unet import Unet
net = Unet(n_channel_in=channels, n_channel_out=channels, **model_params)
net.load_state_dict(state)
else:
net = state
net.to(args.device)
torch.no_grad()
net.eval()
for i in range(len(dataset)):
sharp, pro = dataset.get_full(i)
sharp = sharp.to(args.device)
if args.n2s:
print("N2S evening")
sharp = sharp[:, :(sharp.shape[1] - sharp.shape[1] % 16), sharp.shape[2] % 16:]
pro = pro[:, :(pro.shape[1] - pro.shape[1] %16), pro.shape[2] % 16:]
print(sharp.shape)
paths = dataset.paths_grouped[i]
basename = os.path.basename(paths[0])[:-9]
if channels == 2:
assert not np.all(sharp[0].detach().numpy() == sharp[1].detach().numpy())
if args.channelswap:
sharp = torch.stack((sharp[1], sharp[0]))
if pro is not None:
pro = pro.to('cpu').detach().numpy()
prediction = np.zeros_like(pro)
if args.chop > 1:
idx = [int(x) for x in np.linspace(0, sharp.shape[-2], args.chop)] # indexes
assert idx[-1] == sharp.shape[-2]
for begin, end in zip(idx[:-1], idx[1:]):
prediction[:, begin:end, :] = net(sharp[:, begin:end, :].unsqueeze(0)).squeeze(0).detach().numpy()
else:
prediction = net(sharp.unsqueeze(0)).squeeze(0).detach().numpy()
if channels == 2:
if args.channelswap:
prediction = np.stack((prediction[1], prediction[0]), axis=0)
prediction_high = Image.fromarray(((1.0 - prediction[0]) * 65535).astype(np.uint32))
prediction_low = Image.fromarray(((1.0 - prediction[1]) * 65535).astype(np.uint32))
prediction_high.save(os.path.join(args.outdir, basename+"_high.png"))
prediction_low.save(os.path.join(args.outdir, basename+"_low.png"))
else:
prediction_high = Image.fromarray(((1.0 - prediction[0]) * 65535).astype(np.uint32))
prediction_high.save(os.path.join(args.outdir, basename+"_high.png"))
prediction_high.save(os.path.join(args.outdir, basename+"_high.png"))
prediction_low.save(os.path.join(args.outdir, basename+"_low.png"))
if args.save_gt:
if pro is None:
print("There is not GT to save for sharp data")
gt_high = Image.fromarray(((1.0 - pro[0]) * 65535).astype(np.uint32))
gt_low = Image.fromarray(((1.0 - pro[1]) * 65535).astype(np.uint32))
gt_high.save(os.path.join(args.outdir, "gt_"+basename+"_high.png"))
gt_low.save(os.path.join(args.outdir, "gt_"+basename+"_low.png"))
if args.save_input:
sharp = sharp.detach().numpy()
input_high = Image.fromarray(((1.0 - sharp[0]) * 65535).astype(np.uint32))
input_low = Image.fromarray(((1.0 - sharp[1]) * 65535).astype(np.uint32))
input_high.save(os.path.join(args.outdir, "input_"+basename+"_high.png"))
input_low.save(os.path.join(args.outdir, "input_"+basename+"_low.png"))
if not args.noconvert:
print("generated images, running high low conversion")
os.system("yes | ~/uni/cv-project/cvlab/hilo_converter_v1.2 {} {}".format(args.outdir, args.outdir))