-
Notifications
You must be signed in to change notification settings - Fork 2
/
train.py
208 lines (171 loc) · 8.17 KB
/
train.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
import os
import os.path as osp
import time
import math
from datetime import timedelta
from argparse import ArgumentParser
import glob
from pathlib import Path
import re
import json
from detect import detect
import torch
import cv2
from torch import cuda
from torch.utils.data import DataLoader
from torch.optim import lr_scheduler
from dataset import CosineAnnealingWarmUpRestarts
from tqdm import tqdm
from east_dataset import EASTDataset
from dataset import SceneTextDataset
from model import EAST
import wandb
from deteval import calc_deteval_metrics
def parse_args():
parser = ArgumentParser()
# Conventional args
parser.add_argument('--data_dir', type=str,
default=os.environ.get('SM_CHANNEL_TRAIN', '/opt/ml/input/data/ICDAR17_Korean'))
parser.add_argument('--model_dir', type=str, default=os.environ.get('SM_MODEL_DIR',
'trained_models'))
parser.add_argument('--device', default='cuda' if cuda.is_available() else 'cpu')
parser.add_argument('--num_workers', type=int, default=4)
parser.add_argument('--image_size', type=int, default=1024)
parser.add_argument('--input_size', type=int, default=512)
parser.add_argument('--batch_size', type=int, default=12)
parser.add_argument('--learning_rate', type=float, default=1e-3)
parser.add_argument('--max_epoch', type=int, default=200)
parser.add_argument('--save_interval', type=int, default=5)
parser.add_argument('--exp_name', type=str, default='test') ### wandb 실험 제목, pth 저장하는 폴더 이름
parser.add_argument('--CosineAnealing', type=bool, default=False)
parser.add_argument('--validation', type=bool, default=False)
parser.add_argument('--pretrained_path', type=str, default='')
args = parser.parse_args()
if args.input_size % 32 != 0:
raise ValueError('`input_size` must be a multiple of 32')
return args
def increment_path(model_dir, exp_name, exist_ok=False): ### Custom : 실험명 자동 네이밍 기능
""" Automatically increment path, i.e. trained_models/exp --> trained_models/exp0, trained_models/exp1 etc.
Args:
exist_ok (bool): whether increment path (increment if False).
"""
path = osp.join(model_dir, exp_name)
path = Path(path)
if (path.exists() and exist_ok) or (not path.exists()):
return str(exp_name)
else:
dirs = glob.glob(f"{path}*")
matches = [re.search(rf"%s(\d+)" % path.stem, d) for d in dirs]
i = [int(m.groups()[0]) for m in matches if m]
n = max(i) + 1 if i else 2
return f"{exp_name}{n}"
def min_max_bbox(bbox):
x_list, y_list = [], []
for point in bbox:
x_list.append(point[0])
y_list.append(point[1])
return [min(x_list), min(y_list), max(x_list), max(y_list)]
def do_training(data_dir, model_dir, device, image_size, input_size, num_workers, batch_size,
learning_rate, max_epoch, save_interval, exp_name, CosineAnealing, validation, pretrained_path):
wandb.login()
exp_name = increment_path(model_dir, exp_name)
config = args.__dict__
config['exp_name'] = exp_name
wandb.init(project="myeongu-data", entity="cv-3-bitcoin",name=exp_name, config=config)
### project명 수정 필요합니다!
if validation: ## validation while training
print('Validation is applied!')
dataset = SceneTextDataset(data_dir, split='train_v1', image_size=image_size, crop_size=input_size)
dataset = EASTDataset(dataset)
else:
dataset = SceneTextDataset(data_dir, split='train', image_size=image_size, crop_size=input_size)
dataset = EASTDataset(dataset)
## validation
if validation:
with open(osp.join(data_dir, 'ufo/{}.json'.format('valid_v1')), 'r') as f:
val_anno = json.load(f)
val_image_fname = sorted(val_anno['images'].keys())
val_image_dir = osp.join(data_dir, 'images')
gt_bboxes_dict = {}
for image_fname in val_image_fname:
words_info = val_anno['images'][image_fname]['words'].values()
words_bboxes = [word_info['points'] for word_info in words_info]
gt_bboxes_dict[image_fname] = [min_max_bbox(bbox) for bbox in words_bboxes]
# if image_fname in val_anno['images']:
# print(gt_bboxes_dict)
num_batches = math.ceil(len(dataset) / batch_size)
train_loader = DataLoader(dataset, batch_size=batch_size, shuffle=True, num_workers=num_workers)
device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
model = EAST()
if pretrained_path:
print("pretrained mode!")
model.load_state_dict(torch.load(pretrained_path))
model.to(device)
if CosineAnealing:
optimizer = torch.optim.Adam(model.parameters(), lr = 0)
scheduler = CosineAnnealingWarmUpRestarts(optimizer, T_0=max_epoch//7, T_mult=2, eta_max=0.1, T_up=10, gamma=0.1)
print("[Scheduler]: CosineAnnealingWarmUpRestarts is applied!")
else:
optimizer = torch.optim.Adam(model.parameters(), lr=learning_rate)
scheduler = lr_scheduler.MultiStepLR(optimizer, milestones=[max_epoch // 2], gamma=0.1)
for epoch in range(max_epoch):
model.train()
epoch_loss, epoch_start = 0, time.time()
with tqdm(total=num_batches) as pbar:
for img, gt_score_map, gt_geo_map, roi_mask in train_loader:
pbar.set_description('[Epoch {}]'.format(epoch + 1))
loss, extra_info = model.train_step(img, gt_score_map, gt_geo_map, roi_mask)
optimizer.zero_grad()
loss.backward()
optimizer.step()
loss_val = loss.item()
epoch_loss += loss_val
pbar.update(1)
val_dict = {
'Cls loss': extra_info['cls_loss'], 'Angle loss': extra_info['angle_loss'],
'IoU loss': extra_info['iou_loss']
}
wandb.log({
'Cls loss': extra_info['cls_loss'], 'Angle loss': extra_info['angle_loss'],
'IoU loss': extra_info['iou_loss']
})
### wandb 로깅 부분
pbar.set_postfix(val_dict)
## validation 추가
if validation:
model.eval()
pred_bboxes_dict = {}
val_images = []
pred_bboxes = []
with torch.no_grad():
for image_fname in val_image_fname:
image_fpath = osp.join(val_image_dir, image_fname)
val_images.append(cv2.imread(image_fpath)[:, :, ::-1])
if len(val_images) == batch_size:
pred_bboxes.extend(detect(model, val_images, input_size))
val_images = []
if len(val_images):
pred_bboxes.extend(detect(model, val_images, input_size))
for image_fname, bboxes in zip(val_image_fname, pred_bboxes):
pred_bboxes_dict[image_fname] = [min_max_bbox(bbox) for bbox in bboxes.tolist()]
deteval_metrics = calc_deteval_metrics(pred_bboxes_dict, gt_bboxes_dict)['total']
print('[precision=' + str(deteval_metrics['precision'])
+ ' recall=' + str(deteval_metrics['precision'])
+ ' hmean=' + str(deteval_metrics['hmean']) + '] \n'
)
wandb.log(deteval_metrics)
#################
scheduler.step()
print('Mean loss: {:.4f} | Elapsed time: {}'.format(
epoch_loss / num_batches, timedelta(seconds=time.time() - epoch_start)))
if (epoch + 1) % save_interval == 0:
exp_path = osp.join(model_dir, exp_name)
if not osp.exists(exp_path):
os.makedirs(exp_path)
ckpt_fpath = osp.join(exp_path, 'latest.pth')
torch.save(model.state_dict(), ckpt_fpath)
def main(args):
do_training(**args.__dict__)
if __name__ == '__main__':
args = parse_args()
main(args)