-
Notifications
You must be signed in to change notification settings - Fork 0
/
inference.py
212 lines (156 loc) · 7.2 KB
/
inference.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
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
from pathlib import Path
import torch
from torch.utils.data import DataLoader
from config.defaults import get_cfg_defaults
import time
import numpy as np
import yacs.config
from typing import Dict
from tqdm import tqdm
import pandas as pd
import nibabel as nib
from data.data_utils import sagittal_transform_axial, sagittal_transform_coronal
from models.networks import build_model
from data.dataset import data_set_from_origdata
from utils import wm_logger
logger = wm_logger.loggen(__name__)
class Inference:
def __init__(self,
params: Dict,
cfg: yacs.config.CfgNode
):
self.dataset_path = Path(params["dataset_path"])
self.volume_name = params["volume_name"]
self.volume2_name = params["volume2_name"]
self.volume3_name = params["volume3_name"]
self.csv_file = Path(params["csv_file"])
self.plane = params["plane"]
self.datatype = "test"
self.out_dir = Path(params["out_path"])
self.ckpt = params["ckpt_path"]
assert self.dataset_path.is_dir(), f"The provided paths are not valid: {self.dataset_path}!"
dataframe = pd.read_csv(self.csv_file)
dataframe = dataframe[dataframe["wmh_split"].str.contains(self.datatype)]
self.subjects_dirs = [ self.dataset_path / row["imageid"] for index, row in dataframe.iterrows() ]
logger.info(f"Loading {len(self.subjects_dirs)} for inference")
self.device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
logger.info(f"Using device: {self.device} for inference")
np.random.seed(cfg.RNG_SEED)
torch.manual_seed(cfg.RNG_SEED)
self.cfg = cfg
self.data_set_size = len(dataframe)
self.model = build_model(cfg)
model_state = torch.load(self.ckpt, map_location=self.device)
self.model.load_state_dict(model_state["model_state"])
self.model = self.model.to(self.device)
logger.info(f"Loading {self.ckpt} with model {cfg.MODEL.MODEL_NAME}")
@torch.no_grad()
def get_prediction(self):
self.model.eval()
for idx, current_subject in enumerate(self.subjects_dirs):
start = time.time()
logger.info(
f"Volume Nr: {idx + 1} getting inference from {current_subject.name}/{self.volume_name}"
)
if self.cfg.MODEL.NUM_CHANNELS == 14:
list_input = [current_subject / self.volume_name,
current_subject / self.volume2_name]
dataset = data_set_from_origdata(list_input, self.cfg)
elif self.cfg.MODEL.NUM_CHANNELS == 21:
list_input = [current_subject / self.volume_name,
current_subject / self.volume2_name,
current_subject / self.volume3_name]
dataset = data_set_from_origdata(list_input, self.cfg)
else:
list_input = [current_subject / self.volume_name]
dataset = data_set_from_origdata(list_input,self.cfg)
dataloader = DataLoader(
dataset=dataset,
batch_size=self.cfg.TEST.BATCH_SIZE,
num_workers=self.cfg.TRAIN.NUM_WORKERS,
shuffle=False,
pin_memory=True)
index = 0
out = np.zeros([256,256,256])
out2 = np.zeros([256,2,256,256])
for curr_iter, batch in tqdm(enumerate(dataloader), total=len(dataloader)):
images = batch["image"].to(self.device)
if self.cfg.MODEL.NUM_CHANNELS == 14:
images2 = batch["image2"].to(self.device)
inputs = torch.cat((images, images2), dim=1)
elif self.cfg.MODEL.NUM_CHANNELS == 21:
images2 = batch["image2"].to(self.device)
images3 = batch["image3"].to(self.device)
inputs = torch.cat((images, images2, images3), dim=1)
else:
inputs = images
pred = self.model(inputs)
pred_classes = torch.argmax(pred,dim=1)
prob = torch.nn.functional.softmax(pred, dim=1)
new_index = index+self.cfg.TEST.BATCH_SIZE
out[index:new_index, :, :] = pred_classes.cpu().numpy()
out2[index:new_index, :, :,:] = prob.cpu().numpy()
index = new_index
if self.plane == "axial":
volume = sagittal_transform_axial(out,inverse=True)
elif self.plane == "coronal":
volume = sagittal_transform_coronal(out,inverse=True)
else:
volume = out
affine, header = dataset.get_img_info()
new_img = nib.Nifti1Image(np.asarray((volume > 0.0) * 1.0,np.uint8), affine, header)
logger.info(f"Saving {self.out_dir} / f'{current_subject.name}.nii.gz'")
nib.save(new_img, self.out_dir / f'{current_subject.name}.nii.gz')
logger.info(f"Saving {self.out_dir} / f'{current_subject.name}.npz'")
np.savez(self.out_dir / f'{current_subject.name}.npz',softmax=out2)
logger.info(
"Inference for {} took: {:.3f} seconds".format(
current_subject.name, time.time() - start
))
if __name__ == "__main__":
import argparse
parser = argparse.ArgumentParser("Inference algorithm")
parser.add_argument("--dataset_path",
type=str,
help="name of the dataset to load for inference")
parser.add_argument("--volume_name",
type=str,
help="name of the volume to load for inference")
parser.add_argument("--volume2_name",
type=str,
help="name of the volume to load for inference",
default=None)
parser.add_argument("--volume3_name",
type=str,
help="name of the volume to load for inference",
default=None)
parser.add_argument("--cfg_file",
type=str,
help="path for the cfg file")
parser.add_argument("--csv_file",
type=str,
help="path for the csv file describing the dataset")
parser.add_argument("--plane",
type=str,
help="name of the plane to load in inference")
parser.add_argument("--out_path",
type=str,
help="outpath to save the volumes")
parser.add_argument("--ckpt_path", type=str, help="path to ckpt to load")
parser.add_argument("--num_channels", type=int, help="channels for the input")
args = parser.parse_args()
dataset_params = {
"dataset_path": args.dataset_path,
"volume_name":args.volume_name,
"volume2_name": args.volume2_name,
"volume3_name": args.volume3_name,
"csv_file":args.csv_file,
"plane":args.plane,
"out_path":args.out_path,
"ckpt_path": args.ckpt_path
}
cfg = get_cfg_defaults()
cfg.merge_from_file(args.cfg_file)
cfg.MODEL.NUM_CHANNELS = args.num_channels
inf = Inference(params=dataset_params, cfg=cfg)
inf.get_prediction()