-
Notifications
You must be signed in to change notification settings - Fork 0
/
utils.py
151 lines (120 loc) · 4.95 KB
/
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
#!/usr/bin/python
# -*- coding: utf-8 -*-
from torch.utils.data import Dataset, DataLoader
import os
from PIL import Image
from torchvision import transforms
from torch.autograd import Variable
import torch
import numpy as np
import matplotlib.pyplot as plt
from mpl_toolkits.axes_grid1 import ImageGrid
def imshow(axis, inp):
"""Denormalize and show"""
inp = inp.numpy().transpose((1, 2, 0))
mean = np.array([0.485, 0.456, 0.406])
std = np.array([0.229, 0.224, 0.225])
inp = std * inp + mean
axis.imshow(inp)
class DogDataset(Dataset):
"""Dog Breed Dataset"""
def __init__(self, filenames, labels, root_dir, transform=None):
self.filenames = filenames
self.labels = labels
self.root_dir = root_dir
self.transform = transform
def __len__(self):
return len(self.filenames)
def __getitem__(self, item):
label = self.labels[item]
img_name = os.path.join(self.root_dir, self.filenames[item] + '.jpg')
with Image.open(img_name) as f:
img = f.convert('RGB')
if self.transform:
img = self.transform(img)
if self.labels is None:
return img, self.filenames[item]
else:
return img, self.labels[item]
def get_train_dataset(filenames, labels, batch_size, rootdir='data/train'):
composed = transforms.Compose([transforms.RandomResizedCrop(224, scale=(0.75, 1)),
transforms.RandomHorizontalFlip(),
transforms.ToTensor(),
transforms.Normalize(mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225])
])
dog_trainset = DogDataset(filenames, labels, transform=composed, root_dir=rootdir)
dog_train = DataLoader(dog_trainset, batch_size, True)
return dog_train
def get_test_dataset(filenames, batch_size, rootdir='data/test'):
composed = transforms.Compose([transforms.Resize(224),
transforms.ToTensor(),
transforms.Normalize(mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225])
])
dog_testset = DogDataset(filenames, None, transform=composed, root_dir=rootdir)
dog_test = DataLoader(dog_testset, batch_size, False)
return dog_test
def train_epoch(net, data_iter, criterion, optimizer, use_cuda, print_every=20):
net.eval() # net.train()
correct = 0
for batch_idx, (x, y) in enumerate(data_iter):
if use_cuda:
x, y = x.cuda(), y.cuda()
x = Variable(x)
y = Variable(y)
optimizer.zero_grad()
logits = net(x)
loss = criterion(logits, y)
loss.backward()
optimizer.step()
prediction = torch.argmax(logits, 1)
cur_correct = (prediction == y).sum().float()
cur_accuracy = cur_correct / x.shape[0]
correct += cur_correct
if batch_idx % print_every == 0:
print('current batch: {}/{} ({:.0f}%)\tLoss: {:.6f}\tAcc: {:.6f}'.format(
batch_idx, len(data_iter),
100. * batch_idx / len(data_iter), loss.data.item(), cur_accuracy))
accuracy = correct / len(data_iter.dataset)
print('Train epoch Acc: {:.6f}'.format(accuracy))
return accuracy
def val_epoch(net, data_iter, criterion, use_cuda):
test_loss = 0
correct = 0
net.eval()
for batch_idx, (x, y) in enumerate(data_iter):
if use_cuda:
x, y = x.cuda(), y.cuda()
x = Variable(x)
y = Variable(y)
logits = net(x)
loss = criterion(logits, y)
test_loss += loss.data.item()
prediction = torch.argmax(logits, 1)
cur_correct = (prediction == y).sum().float()
correct += cur_correct
test_loss /= len(data_iter.dataset)
accuracy = correct / len(data_iter.dataset)
print('\nVal set: Average loss: {:.4f}, Accuracy: {}/{} ({:.0f}%)\n'
.format(test_loss, correct, len(data_iter.dataset), 100. * accuracy))
return accuracy
def visualize_model(dataloders, model, num_images=16, use_gpu=True):
cnt = 0
fig = plt.figure(1, figsize=(16, 16))
grid = ImageGrid(fig, 111, nrows_ncols=(4, 4), axes_pad=0.05)
for i, (inputs, labels) in enumerate(dataloders['valid']):
if use_gpu:
inputs, labels = Variable(inputs.cuda()), Variable(labels.cuda())
else:
inputs, labels = Variable(inputs), Variable(labels)
outputs = model(inputs)
_, preds = torch.max(outputs.data, 1)
for j in range(inputs.size()[0]):
ax = grid[cnt]
imshow(ax, inputs.cpu().data[j])
ax.text(10, 210, '{}/{}'.format(preds[j], labels.data[j]),
color='k', backgroundcolor='w', alpha=0.8)
cnt += 1
if cnt == num_images:
return