/
train.py
208 lines (168 loc) · 8.26 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
# -*- coding: utf-8 -*-
import argparse
import os
# os.environ["CUDA_VISIBLE_DEVICES"] = '1'
import torch
import torch.nn as nn
import torch.optim as optim
from torch.utils.data import DataLoader
import torchvision
from torchvision import transforms
# from data_loader import ImageData
from models import resnet, efficientnet, resnest, resnext, mymodel
from utils import auto_augment, losses, data_loader
import pandas as pd
import config
import json
from tqdm import tqdm
cfg = config.Config()
if not os.path.exists(cfg.model_path):
os.mkdir(cfg.model_path)
# whether use gpu
use_gpu = torch.cuda.is_available()
if use_gpu:
DEVICE = torch.device('cuda')
else:
DEVICE = torch.device('cpu')
class Main():
def data_process(self):
normal = transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])
if cfg.use_autoaugment:
train_transform = transforms.Compose([
transforms.Resize(cfg.image_size),
transforms.CenterCrop(cfg.crop_image_size),
# transforms.RandomRotation(cfg.rotate_degree),
auto_augment.AutoAugment(dataset=cfg.autoaugment_dataset),
transforms.ToTensor(),
transforms.RandomErasing(p=0.3, scale=(0.02, 0.1), ratio=(0.6, 1.2)),
normal
])
else:
train_transform = transforms.Compose([
transforms.Resize(cfg.image_size),
transforms.CenterCrop(cfg.crop_image_size),
transforms.RandomRotation(cfg.rotate_degree),
transforms.ToTensor(),
transforms.RandomErasing(p=0.3, scale=(0.02, 0.1), ratio=(0.6, 1.2)),
normal
])
test_trainsform = transforms.Compose([
transforms.Resize(cfg.image_size),
transforms.CenterCrop(cfg.crop_image_size),
transforms.ToTensor(),
normal
])
return train_transform, test_trainsform
def train_model(self, model, epochs, train_loader, val_loader, save_model_name):
model.to(DEVICE)
if cfg.use_label_smooth:
criterion = losses.CrossEntropyLabelSmooth(cfg.num_class, epsilon=0.1)
else:
criterion = nn.CrossEntropyLoss()
criterion.to(DEVICE)
if cfg.optim == 'adam':
optimizer = optim.Adam(model.parameters(), lr=cfg.lr, weight_decay=cfg.weight_decay_rate)
else:
optimizer = optim.SGD(model.parameters(), lr=cfg.lr, momentum=cfg.momentum_rate, weight_decay=cfg.weight_decay_rate, nesterov=cfg.use_nesterov)
scheduler = optim.lr_scheduler.CosineAnnealingLR(optimizer, T_max=epochs, eta_min=1e-5)
for epoch in range(epochs):
model.train()
train_loss = 0
for i, (img, label) in enumerate(train_loader):
img, label = img.to(DEVICE), label.to(DEVICE)
# print(img.size())
# print(label.size())
optimizer.zero_grad()
# output = model(img)
c0, c1, c2, c3, c4 = model(img)
loss = criterion(c0, label[:, 0]) + \
criterion(c1, label[:, 1]) + \
criterion(c2, label[:, 2]) + \
criterion(c3, label[:, 3]) + \
criterion(c4, label[:, 4])
# loss = criteration(output, label)
train_loss += loss.item()
loss.backward()
optimizer.step()
if i % 100 == 0:
print("Epoch {}, Step {}, Train_Loss {:.4f}".format(epoch, i , train_loss))
train_loss = 0
scheduler.step()
if epoch % 1 == 0 or epoch == epochs - 1:
correct = 0
model.eval()
for val_img, val_label in tqdm(val_loader):
val_img, val_label = val_img.to(DEVICE), val_label.to(DEVICE)
p1, p2, p3, p4, p5 = model(val_img)
val_pred1 = torch.max(p1, 1)[1]
val_pred2 = torch.max(p2, 1)[1]
val_pred3 = torch.max(p3, 1)[1]
val_pred4 = torch.max(p4, 1)[1]
val_pred5 = torch.max(p5, 1)[1]
# print(val_pred1)
# print(val_pred2)
# print(val_pred3)
# print(val_pred4)
# print(val_pred5)
val_pred = torch.stack([val_pred1, val_pred2, val_pred3, val_pred4, val_pred5], 1)
for i in range(val_pred.size(0)):
flag = True
for j in range(val_pred[i].size(0)):
if val_pred[i][j].item() != val_label[i][j].item():
flag = False
if flag:
correct += 1
print("Epoch {}, Accuracy {:.2f}%".format(epoch, 100 * correct / len(val_loader.dataset)))
# torch.save(model, cfg.model_path + '/epoch_' + str(epoch) + '.pth')
torch.save(model, cfg.model_path + '/' + save_model_name)
print(f'Current model-{save_model_name} ends, saved the model.')
def train(self):
# df = pd.read_csv(os.path.join(cfg.data_path, 'train.csv'))
# image_path_list = df['image_path'].values
# label_list = df['label'].values
# split dataset
# all_size = len(image_path_list)
# train_size = int(all_size * cfg.train_per)
# train_image_path_list = image_path_list[:train_size]
# train_label_list = label_list[:train_size]
# val_image_path_list = image_path_list[train_size:]
# val_label_list = label_list[train_size:]
# print(
# 'train_size: %d, val_size: %d' % (len(train_image_path_list), len(val_image_path_list)))
train_image_path_list = os.listdir(os.path.join(cfg.data_path, 'mchar_train'))
train_image_path_list.sort()
train_image_path_list = [os.path.join('mchar_train', name) for name in train_image_path_list]
with open(os.path.join(cfg.data_path, 'mchar_train.json')) as t_f:
train_labels = json.load(t_f)
train_label_list = [train_labels[img]['label'] for img in train_labels.keys()]
val_image_path_list = os.listdir(os.path.join(cfg.data_path, 'mchar_val'))
val_image_path_list.sort()
val_image_path_list = [os.path.join('mchar_val', name) for name in val_image_path_list]
with open(os.path.join(cfg.data_path, 'mchar_val.json')) as v_f:
val_labels = json.load(v_f)
val_label_list = [val_labels[img]['label'] for img in val_labels.keys()]
train_transform, val_transform = self.data_process()
train_data = data_loader.ImageData(train_image_path_list, train_label_list, train_transform, cfg.data_path)
val_data = data_loader.ImageData(val_image_path_list, val_label_list, val_transform, cfg.data_path)
# filename = '/home/data/14'
# all_data = torchvision.datasets.ImageFolder(filename, transform=train_transform)
# trainsize = int(0.9 * len(all_data))
# valsize = len(all_data) - trainsize
# train_data, val_data = torch.utils.data.random_split(all_data,[trainsize, valsize])
# print(all_data.class_to_idx)
# print(train_data.class_to_idx)
train_loader = DataLoader(train_data, batch_size=cfg.train_batch_size, shuffle=True, num_works = cfg.num_worker, drop_last=True)
val_loader = DataLoader(val_data, batch_size=cfg.val_batch_size, shuffle=False, num_works = cfg.num_worker, drop_last=True)
# model1 = mymodel.MyModel('resnest101', cfg.pretrained, cfg.is_local, cfg.num_class)
# print(model1)
# print('Loading model1...')
# self.train_model(model1, cfg.epochs, train_loader, val_loader, 'best1.pth')
model2 = mymodel.MyModel2('efficientnet-b6', cfg.pretrained, cfg.is_local, cfg.num_class)
if torch.cuda.device_count() > 1:
model2 = torch.nn.DataParallel(model2, device_ids=cfg.device_id)
print(model2)
# print('Loading model2...')
self.train_model(model2, cfg.epochs, train_loader, val_loader, 'best2.pth')
if __name__ == '__main__':
main = Main()
main.train()