-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathmain.py
149 lines (124 loc) · 5.26 KB
/
main.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
from model import *
from datasets import *
from train import *
from evaluate import *
import transforms as T
import numpy as np
from utils import *
from sklearn.preprocessing import OneHotEncoder, LabelEncoder
args = parser.parse_args()
print("Args: ", args)
# Dataloader parameters
batch_size = 40
image_height, image_width = 112, 112 # resize video 2d frame size
n_frames = 15 # number of frames in a video clip
fps = 10
random_slice_size = 0
num_classes = 4
categories = [0, 1, 2, 3]
# Detect devices
use_cuda = torch.cuda.is_available() # check if GPU exists
if use_cuda:
print("============== USING CUDA ==============")
params = {'batch_size': batch_size, 'shuffle': True, 'pin_memory': True}
device = torch.device("cuda") # use CPU or GPU
else:
print("============== USING CPU ==============")
device = torch.device("cpu")
params = {'batch_size': batch_size, 'shuffle': True, 'pin_memory': True}
train_list, train_label = read_data_labels('train1.csv', categories)
test_list, test_label = read_data_labels('valid1.csv', categories)
if args.crop_videos:
crop_video(train_list, train_label)
crop_video(test_list, test_label)
categories = [0, 1, 2, 3]
if args.ordinal:
# convert category to multi-hot
label_enc = LabelEncoder()
label_enc.fit(categories)
action_category = label_enc.transform(categories).reshape(-1, 1)
enc = OneHotEncoder()
t = enc.fit(action_category)
train_label = train_label.reshape(-1, 1)
train_label = np.cumsum(enc.transform(train_label).toarray(), axis=1)[:, 0:3]
test_label = test_label.reshape(-1, 1)
test_label = np.cumsum(enc.transform(test_label).toarray(), axis=1)[:, 0:3]
spatial_transform_train = torchvision.transforms.Compose([
T.ToFloatTensorInZeroOne(),
T.Resize((image_height, image_width)),
T.RandomHorizontalFlip(),
# Normalization done after data is loaded
])
spatial_transform_test = torchvision.transforms.Compose([
T.ToFloatTensorInZeroOne(),
T.Resize((image_height, image_width)),
# Normalization done after data is loaded
])
print("============== Loading Data ==============")
print("Train {} videos".format(len(train_list)))
print("Test {} videos".format(len(test_list)))
train_set = MyVideoDataset('./new_video_data_clip', train_list, train_label, n_frames=n_frames, fps=fps, spatial_transform=spatial_transform_train, random_slice_size=random_slice_size)
valid_set = MyVideoDataset('./new_video_data_clip', test_list, test_label, n_frames=n_frames, fps=fps, spatial_transform=spatial_transform_test, random_slice_size=random_slice_size)
train_loader = data.DataLoader(train_set, **params)
valid_loader = data.DataLoader(valid_set, **params)
print("Train {} clips".format(len(train_set)))
print("Test {} clips".format(len(valid_set)))
# Normalize Data
if args.get_stats:
m_, s_ = get_stats(train_loader)
print("Calculated stats: mean ", m_, "and std ", s_)
else:
if fps == 10:
m_ = torch.tensor([0.5078, 0.4929, 0.4816])
s_ = torch.tensor([0.2329, 0.2376, 0.2498])
else: # fps == 1
m_ = torch.tensor([0.4926, 0.4835, 0.4777])
s_ = torch.tensor([0.2355, 0.2401, 0.2485])
# create model
if args.pretrained:
model = Conv3dModelPretrained(num_classes=num_classes).to(device)
elif args.ordinal:
model = OrdinalModelPretrained(num_classes=num_classes).to(device)
else:
model = Conv3dModel(image_t_frames=n_frames, image_height=image_height, image_width=image_width, num_classes=num_classes).to(device)
print("Model: ", model)
# Parallelize model to multiple GPUs
if torch.cuda.device_count() > 1:
print("============== USING", torch.cuda.device_count(), "GPUs ==============")
model = nn.DataParallel(model)
# training parameters
epochs = 10
learning_rate = 1e-4
optimizer = torch.optim.Adam(model.parameters(), lr=learning_rate) # optimize all cnn parameters
# record training process
epoch_train_losses = []
epoch_train_scores = []
epoch_test_losses = []
epoch_test_scores = []
pre_epoch = 0
if args.resume:
print('resume--- ')
checkpoint = torch.load('last.pth')
model.load_state_dict(checkpoint['model_state_dict'])
optimizer.load_state_dict(checkpoint['optimizer_state_dict'])
pre_epoch = checkpoint['epoch'] + 1
loss = checkpoint['loss']
print('epoch ')
# start training
for epoch in range(pre_epoch, epochs):
# train, test model
train_losses, train_scores = train_one_epoch(model, device, train_loader, optimizer, epoch)
epoch_test_loss, epoch_test_score = evaluate(model, device, valid_loader)
# save results
epoch_train_losses.append(train_losses)
epoch_train_scores.append(train_scores)
epoch_test_losses.append(epoch_test_loss)
epoch_test_scores.append(epoch_test_score)
print("Train losses: ", np.array(epoch_train_losses))
print("Train scores: ", np.array(epoch_train_scores))
print("Test losses: ", np.array(epoch_test_losses))
print("Test scores: ", np.array(epoch_test_scores))
# np.save('./3DCNN_epoch_{}_training_losses.npy'.format(epoch), np.array(epoch_train_losses))
# np.save('./3DCNN_epoch_{}_training_scores.npy'.format(epoch), np.array(epoch_train_losses))
# np.save('./3DCNN_epoch_{}_test_loss.npy'.format(epoch), np.array(epoch_test_losses))
# np.save('./3DCNN_epoch_{}_test_score.npy'.format(epoch), np.array(epoch_test_scores))