Skip to content

Navigation Menu

Sign in
Appearance settings

Search code, repositories, users, issues, pull requests...

Provide feedback

We read every piece of feedback, and take your input very seriously.

Saved searches

Use saved searches to filter your results more quickly

Appearance settings

Latest commit

 

History

History
History
245 lines (202 loc) · 7.56 KB

File metadata and controls

245 lines (202 loc) · 7.56 KB
Copy raw file
Download raw file
Open symbols panel
Edit and raw actions
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
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
from pathlib import Path
import glob
import torch
import numpy as np
from models import *
from utils.helper_funcs import (
load_config,
save_sampling_results_as_imgs,
get_model_path,
get_conf_name,
print_config,
draw_boundary,
mean_of_list_of_tensors,
)
from forward.forward_schedules import ForwardSchedule
from reverse.reverse_process import sample
import matplotlib.pyplot as plt
from tqdm import tqdm
from modules.transforms import DiffusionTransform
import torchvision
import json
from PIL import Image
import shutil
from torch.utils.tensorboard import SummaryWriter
from common.logging import get_logger
from argument import get_argparser, sync_config
import sys, os
from metrics import get_binary_metrics
from loaders.dataloaders import get_dataloaders
import warnings
warnings.filterwarnings("ignore")
# ------------------- params --------------------
argparser = get_argparser()
args = argparser.parse_args(sys.argv[1:])
config = load_config(args.config_file)
config = sync_config(config, args)
logger = get_logger(
filename=f"{config['model']['name']}_test", dir=f"logs/{config['dataset']['name']}"
)
print_config(config, logger)
jet = plt.get_cmap("jet")
def write_imgs(
imgs, msks, prds, mid_prds, step, id, dataset, ids=None
):
imgs = (imgs - imgs.min()) / (imgs.max() - imgs.min())
img_grid = torchvision.utils.make_grid(imgs)
msk_grid = torchvision.utils.make_grid(msks)
prd_grid = torchvision.utils.make_grid(prds)
mid_prds_jet = torch.zeros_like(imgs)
for i, mid_prd in enumerate(mid_prds.detach().cpu().numpy()):
t = jet(mid_prd[0]).transpose(2, 0, 1)[:-1, :, :]
t = np.log(t + 0.1)
t = (t - t.min()) / (t.max() - t.min())
mid_prds_jet[i, :, :, :] = torch.tensor(t)
mid_prd_grid = torchvision.utils.make_grid(mid_prds_jet)
res_grid = draw_boundary(torch.where(msk_grid > 0, 1, 0), img_grid, (0, 255, 0))
res_grid = draw_boundary(torch.where(prd_grid > 0, 1, 0), res_grid, (0, 0, 255))
img_msk_prd_grid = torch.concat(
[
img_grid,
# torch.stack(3*[msk_grid[0],],0),
mid_prd_grid,
torch.tensor(res_grid).to(device)
# torch.stack(3*[prd_grid[0],], 0),
# torch.stack(3*[prd_grid[0]>0,],0),
],
dim=1,
)
# writer.add_image(f'ISIC 2018 - Results/{"-".join(ids)}', img_msk_prd_grid)
writer.add_image(f"{dataset}/Test:{id}", img_msk_prd_grid, step)
writer = SummaryWriter(f'{config["run"]["writer_dir"]}/{config["model"]["name"]}')
timesteps = config["diffusion"]["schedule"]["timesteps"]
epochs = config["training"]["epochs"]
INPUT_SIZE = config["dataset"]["input_size"]
batch_size = config["data_loader"]["train"]["batch_size"]
img_channels = config["dataset"]["img_channels"]
msk_channels = config["dataset"]["msk_channels"]
ensemble = config["testing"]["ensemble"]
ID = get_conf_name(config)
device = torch.device(config["run"]["device"])
# device = "cuda" if torch.cuda.is_available() else "cpu"
logger.info(f"Device is <{device}>")
# <<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<
# --------- check required dirs --------------------
Path(config["model"]["save_dir"]).mkdir(exist_ok=True)
if config["testing"]["result_imgs"]["save"]:
try:
Path(config["testing"]["result_imgs"]["dir"] + "/" + ID).mkdir()
except FileExistsError:
shutil.rmtree(Path(config["testing"]["result_imgs"]["dir"] + "/" + ID))
Path(config["testing"]["result_imgs"]["dir"] + "/" + ID).mkdir()
forward_schedule = ForwardSchedule(**config["diffusion"]["schedule"])
DT = DiffusionTransform((INPUT_SIZE, INPUT_SIZE))
# --------------- Datasets and Dataloaders -----------------
te_dataloader = get_dataloaders(config, "te")
Net = globals()[config["model"]["class"]]
model = Net(**config["model"]["params"])
model.to(device)
# ------------------------ EMA -------------------------------
# https://github.com/lucidrains/ema-pytorch
from ema_pytorch import EMA
try:
if config["training"]["ema"]["use"]:
ema = EMA(model=model, **config["training"]["ema"]["params"])
ema.to(device)
else:
ema = None
except KeyError:
logger.exception("You need to determine the EMA parameters at <config.training>!")
# >>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
if config["testing"]["model_weigths"]["overload"]:
logger.info(f"trying to load the desired model:")
best_model_path = config["testing"]["model_weigths"]["file_path"]
else:
logger.info(f"trying to load the best model:")
best_model_path = get_model_path(name=ID, dir=config["model"]["save_dir"])
logger.info(f" -> {best_model_path}")
if not os.path.isfile(best_model_path):
logger.exception(f"wanted to load {best_model_path} file but it does not exist!")
checkpoint = torch.load(best_model_path, map_location="cpu")
try:
if ema:
ema.load_state_dict(checkpoint["ema"])
model = ema.ema_model
logger.info("ema loaded...")
else:
model.load_state_dict(checkpoint["model"])
logger.info("simple model loaded...")
except:
logger.exception("Something happened on loading the model weights!")
if torch.cuda.is_available():
torch.cuda.empty_cache()
model.to(device)
total_params = sum(p.numel() for p in model.parameters())
logger.info(f"Number of model parameters: {total_params}")
model.eval()
test_metrics = get_binary_metrics()
for step, batch in tqdm(
enumerate(te_dataloader),
desc=f"Testing {config['model']['name']}",
total=len(te_dataloader),
):
batch_imgs = batch["image"].to(device)
batch_msks = batch["mask"].to(device)
batch_ids = batch["id"]
samples_list, mid_samples_list = [], []
all_samples_list = []
for en in range(ensemble):
samples = sample(
forward_schedule,
model,
images=batch_imgs,
out_channels=batch_msks.shape[1],
desc=f"ensemble {en+1}/{ensemble}",
)
samples_list.append(samples[-1][:, :1, :, :].to(device))
mid_samples_list.append(
samples[-int(0.1 * timesteps)][:, :1, :, :].to(device)
)
all_samples_list.append([s[:, :1, :, :] for s in samples])
# preds = samples[-1].to(device)
preds = mean_of_list_of_tensors(samples_list)
mid_preds = mean_of_list_of_tensors(mid_samples_list)
if batch_msks.shape[1] > 1:
batch_msks = batch_msks[:, 0, :, :].unsqueeze(1)
if batch_msks.shape[1] > 1:
preds_ = torch.argmax(preds, 1, keepdim=False).float()
msks_ = torch.argmax(batch_msks, 1, keepdim=False)
else:
preds_ = torch.where(preds > 0, 1, 0).float()
msks_ = torch.where(batch_msks > 0, 1, 0)
test_metrics.update(preds_, msks_)
write_imgs(
batch_imgs,
batch_msks,
preds,
mid_preds,
step=step,
id=f"{ID}_BV_E{ensemble}",
dataset=config["dataset"]["name"].upper()
)
if config["testing"]["result_imgs"]["save"]:
save_sampling_results_as_imgs(
batch_imgs,
batch_msks,
batch_ids,
preds,
all_samples_list,
middle_steps_of_sampling=8,
save_dir=config["testing"]["result_imgs"]["dir"],
dataset_name=config["dataset"]["name"].upper(),
result_id=f"{ID}_BV_E{ensemble}",
img_ext="png",
save_mat=True,
)
result = test_metrics.compute()
writer.add_scalars(
f"Metrics/test-s{INPUT_SIZE}/{ID}_BV_E{ensemble}",
result,
)
logger.info(f"result for best model {ID}-E{ensemble}")
logger.info(json.dumps({k: v.item() for k, v in result.items()}, indent=4))
Morty Proxy This is a proxified and sanitized view of the page, visit original site.