forked from ultralytics/yolov5
-
Notifications
You must be signed in to change notification settings - Fork 0
/
patchmentation_utils.py
302 lines (237 loc) · 12.1 KB
/
patchmentation_utils.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
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
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
import yaml
import os, shutil
import matplotlib.pyplot as plt
import pandas as pd
import numpy as np
from typing import List
def _remove_ext(file: str):
return os.path.splitext(file)[0]
def _zip(folder: str, file: str):
os.makedirs(os.path.dirname(file), exist_ok=True)
if os.path.exists(file):
raise FileExistsError(file)
print(f'zip {folder} to {file}')
shutil.make_archive(_remove_ext(file), 'zip', folder)
def _unzip(file: str, folder: str):
os.makedirs(folder)
print(f'unzip {file} to {folder}')
shutil.unpack_archive(file, folder, 'zip')
def _bash_upload(file: str):
print(f'bash upload {file}')
os.system(f'curl bashupload.com -T {file}')
def _bash_download(url: str, file: str):
os.makedirs(os.path.dirname(file), exist_ok=True)
if os.path.exists(file):
raise FileExistsError(file)
print(f'bash download {url} to {file}')
os.system(f'wget -c {url} -O {file}')
def _rm(path: str, check: bool = False):
if check and not os.path.exists(path):
return
print(f'remove {path}')
os.system(f'rm -rf {path}')
def _link(source: str, target: str):
print(f'link {source} to {target}')
os.system(f'ln -s {os.path.abspath(source)} {os.path.abspath(target)}')
def _unlink(path: str, check: bool = False):
if check and not os.path.exists(path):
return
print(f'unlink {path}')
os.system(f'rm {path}')
def get_folder_output(project: str, name: str):
return os.path.join(project, name)
def get_folder_output_test(project: str, name: str):
return os.path.join(get_folder_output(project, name), 'test')
def get_weights(project: str, name: str):
return os.path.join(get_folder_output(project, name), 'weights', 'best.pt')
def get_file_zip(project: str, name: str):
return os.path.join(project, f'{name}.zip')
def remove(project: str, name: str, check: bool = False):
folder_output = get_folder_output(project, name)
_rm(folder_output, check)
def remove_test(project: str, name: str, check: bool = False):
folder_output_test = get_folder_output_test(project, name)
_rm(folder_output_test, check)
def remove_zip(project: str, name: str, check: bool = False):
file_zip = get_file_zip(project, name)
_rm(file_zip, check)
def zip(project: str, name: str):
folder_output = get_folder_output(project, name)
file_zip = get_file_zip(project, name)
_zip(folder_output, file_zip)
def unzip(project: str, name: str):
folder_output = get_folder_output(project, name)
file_zip = get_file_zip(project, name)
_unzip(file_zip, folder_output)
def upload(project: str, name: str):
file_zip = get_file_zip(project, name)
_bash_upload(file_zip)
def download(project: str, name: str, url: str):
file_zip = get_file_zip(project, name)
_bash_download(url, file_zip)
def plot(project: str, name: str):
plot_results_mix(project, name)
def plot_results_mix(project: str, name: str):
csv_results = pd.read_csv(get_file_csv_results(project, name))
plot_mix_train_val(csv_results, get_file_plot_results_mix(project, name))
def plot_mix_train_val_ax(ax, data_train, data_val, title: str = None):
index_train = np.arange(len(data_train))
index_val = np.arange(len(data_val))
ax.set_title(title)
ax.plot(index_train, data_train, color='green', label='train')
ax.plot(index_val, data_val, color='red', label='val')
ax.legend()
ax.grid()
def get_csv_results_index(csv_results, index: str):
for key in csv_results.keys():
if key.strip() == index:
return key
raise Exception(index)
def plot_mix_train_val(csv_results, path: str):
fig, axs = plt.subplots(1, 3, figsize=(30, 6), dpi=150)
train_box_loss = csv_results[get_csv_results_index(csv_results, 'train/box_loss')]
val_box_loss = csv_results[get_csv_results_index(csv_results, 'val/box_loss')]
plot_mix_train_val_ax(axs[0], train_box_loss, val_box_loss, 'box_loss')
train_obj_loss = csv_results[get_csv_results_index(csv_results, 'train/obj_loss')]
val_obj_loss = csv_results[get_csv_results_index(csv_results, 'val/obj_loss')]
plot_mix_train_val_ax(axs[1], train_obj_loss, val_obj_loss, 'obj_loss')
train_cls_loss = csv_results[get_csv_results_index(csv_results, 'train/cls_loss')]
val_cls_loss = csv_results[get_csv_results_index(csv_results, 'val/cls_loss')]
plot_mix_train_val_ax(axs[2], train_cls_loss, val_cls_loss, 'cls_loss')
plt.savefig(path)
def get_file_csv_results(project: str, name: str):
return os.path.join(project, name, 'results.csv')
def get_file_plot_results_mix(project: str, name: str):
return os.path.join(project, name, 'results-mix.png')
def get_file_yaml(name: str) -> str:
return os.path.join('data', f'patchmentation-{name}.yaml')
def get_folder_batch_images(folder: str, batch: int):
return os.path.join(folder, str(batch).zfill(3), 'images')
def get_folder_images(folder: str):
return os.path.join(folder, 'images')
def get_folder_batch_annotations(folder: str, batch: int):
return os.path.join(folder, str(batch).zfill(3), 'labels')
def get_folder_annotations(folder: str):
return os.path.join(folder, 'labels')
def get_file_cache(folder: str):
return os.path.join(folder, 'labels.cache')
def create_folder_link(folder: str, batch: int):
folder_batch_images = get_folder_batch_images(folder, batch)
folder_batch_annotations = get_folder_batch_annotations(folder, batch)
folder_link_images = get_folder_images(folder)
folder_link_annotations = get_folder_annotations(folder)
_link(folder_batch_images, folder_link_images)
_link(folder_batch_annotations, folder_link_annotations)
def remove_folder_link(folder: str, check: bool = False):
folder_link_images = get_folder_images(folder)
folder_link_annotations = get_folder_annotations(folder)
_unlink(folder_link_images, check)
_unlink(folder_link_annotations, check)
def link_dataset(path: str, folders: List[str], batch: int):
for folder in folders:
folder_images = os.path.join(path, folder)
create_folder_link(os.path.dirname(folder_images), batch)
def unlink_dataset(path: str, folders: List[str], check: bool = False):
for folder in folders:
folder_images = os.path.join(path, folder)
remove_folder_link(os.path.dirname(folder_images), check)
def remove_cache(path: str, folders: List[str], check: bool = False):
for folder in folders:
folder_images = os.path.join(path, folder)
file_cache = get_file_cache(os.path.dirname(folder_images))
_rm(file_cache, check)
SINGLE_BATCH = 'single-batch'
MULTI_BATCH = 'multi-batch'
def train(data: str, hyp: str, weights: str, epochs: int, batch_size: int, project: str, name: str):
print(f'=================== TRAIN =====================')
print(f'data: {data}')
print(f'hyp: {hyp}')
print(f'weights: {weights}')
print(f'epochs : {epochs}')
print(f'batch_size: {batch_size}')
print(f'project: {project}')
print(f'name: {name}')
print(f'===============================================')
if os.path.exists(get_folder_output(project, name)):
raise FileExistsError(get_folder_output(project, name))
with open(data, 'r') as f:
yaml_dict = yaml.safe_load(f)
train_format = yaml_dict.get('train-format', None)
val_format = yaml_dict.get('val-format', None)
test_format = yaml_dict.get('test-format', None)
print(f'train_format: {train_format}')
print(f'val_format: {val_format}')
print(f'test_format: {test_format}')
print(f'===============================================')
if train_format is not None and train_format in (SINGLE_BATCH, MULTI_BATCH):
remove_cache(yaml_dict['path'], yaml_dict['train'], check=True)
unlink_dataset(yaml_dict['path'], yaml_dict['train'], check=True)
link_dataset(yaml_dict['path'], yaml_dict['train'], 0)
if val_format is not None and val_format in (SINGLE_BATCH, MULTI_BATCH):
remove_cache(yaml_dict['path'], yaml_dict['val'], check=True)
unlink_dataset(yaml_dict['path'], yaml_dict['val'], check=True)
link_dataset(yaml_dict['path'], yaml_dict['val'], 0)
if test_format is not None and test_format in (SINGLE_BATCH, MULTI_BATCH):
remove_cache(yaml_dict['path'], yaml_dict['test'], check=True)
unlink_dataset(yaml_dict['path'], yaml_dict['test'], check=True)
link_dataset(yaml_dict['path'], yaml_dict['test'], 0)
command_initial = f'python3 train.py --data {data} --hyp {hyp} --weights {weights} --epochs {epochs} --batch-size {batch_size} --project {project} --name {name}'
os.system(command_initial)
resume_weight = os.path.join(project, name, 'weights', 'last.pt')
command_resume = f'python3 train.py --resume {resume_weight}'
for epoch in range(1, epochs):
if train_format is not None and train_format in (MULTI_BATCH):
remove_cache(yaml_dict['path'], yaml_dict['train'])
unlink_dataset(yaml_dict['path'], yaml_dict['train'])
link_dataset(yaml_dict['path'], yaml_dict['train'], epoch)
if val_format is not None and val_format in (MULTI_BATCH):
remove_cache(yaml_dict['path'], yaml_dict['val'])
unlink_dataset(yaml_dict['path'], yaml_dict['val'])
link_dataset(yaml_dict['path'], yaml_dict['val'], epoch)
if test_format is not None and test_format in (MULTI_BATCH):
remove_cache(yaml_dict['path'], yaml_dict['test'])
unlink_dataset(yaml_dict['path'], yaml_dict['test'])
link_dataset(yaml_dict['path'], yaml_dict['test'], epoch)
os.system(command_resume)
if train_format is not None and train_format in (SINGLE_BATCH, MULTI_BATCH):
remove_cache(yaml_dict['path'], yaml_dict['train'])
unlink_dataset(yaml_dict['path'], yaml_dict['train'])
if val_format is not None and val_format in (SINGLE_BATCH, MULTI_BATCH):
remove_cache(yaml_dict['path'], yaml_dict['val'])
unlink_dataset(yaml_dict['path'], yaml_dict['val'])
if test_format is not None and test_format in (SINGLE_BATCH, MULTI_BATCH):
remove_cache(yaml_dict['path'], yaml_dict['test'])
unlink_dataset(yaml_dict['path'], yaml_dict['test'])
def test(data: str, weights:str, batch_size: int, project: str, name: str):
print(f'=================== TEST ======================')
print(f'data: {data}')
print(f'weights: {weights}')
print(f'batch_size: {batch_size}')
print(f'project: {project}')
print(f'name: {name}')
print(f'===============================================')
if os.path.exists(get_folder_output_test(project, name)):
raise FileExistsError(get_folder_output_test(project, name))
with open(data, 'r') as f:
yaml_dict = yaml.safe_load(f)
val_format = yaml_dict.get('val-format', None)
test_format = yaml_dict.get('test-format', None)
print(f'val_format: {val_format}')
print(f'test_format: {test_format}')
print(f'===============================================')
if val_format is not None and val_format in (SINGLE_BATCH, MULTI_BATCH):
remove_cache(yaml_dict['path'], yaml_dict['val'], check=True)
unlink_dataset(yaml_dict['path'], yaml_dict['val'], check=True)
link_dataset(yaml_dict['path'], yaml_dict['val'], 0)
if test_format is not None and test_format in (SINGLE_BATCH, MULTI_BATCH):
remove_cache(yaml_dict['path'], yaml_dict['test'], check=True)
unlink_dataset(yaml_dict['path'], yaml_dict['test'], check=True)
link_dataset(yaml_dict['path'], yaml_dict['test'], 0)
command = f'python3 val.py --data {data} --weights {weights} --batch-size {batch_size} --verbose --task test --project {project} --name {name + "/test"}'
os.system(command)
if val_format is not None and val_format in (SINGLE_BATCH, MULTI_BATCH):
remove_cache(yaml_dict['path'], yaml_dict['val'])
unlink_dataset(yaml_dict['path'], yaml_dict['val'])
if test_format is not None and test_format in (SINGLE_BATCH, MULTI_BATCH):
remove_cache(yaml_dict['path'], yaml_dict['test'])
unlink_dataset(yaml_dict['path'], yaml_dict['test'])