opts_file = open(path.join(logdir, 'opts.txt'), 'wt')
opts_file.write(str(opt))
opts_file.close()
loss_outfile = open(path.join(logdir, 'losses.jsonl'), 'wt')
# Load dataset
if os.path.isfile('celeba.pickle'):
print('loading dataset cached')
dataset = pickle.load(open('celeba.pickle', 'rb'))
else:
print('loading dataset new')
dataset = ImageFolder(root=opt.datadir,
transform=transforms.Compose([
transforms.ToTensor(),
transforms.Normalize((0.5, 0.5, 0.5),
(0.5, 0.5, 0.5)),
]))
print('dumping datset to celeba.pickle')
pickle.dump(dataset, open('celeba.pickle', 'wb'))
dataloader = torch.utils.data.DataLoader(dataset,
batch_size=opt.batchsize,
shuffle=True,
num_workers=int(2))
if opt.manualSeed is None:
opt.manualSeed = random.randint(1, 10000)
print("Random Seed: ", opt.manualSeed)
random.seed(opt.manualSeed)
torch.manual_seed(opt.manualSeed)
scores = []
for i in range(part.shape[0]):
pyx = part[i, :]
scores.append(entropy(pyx, py))
split_scores.append(np.exp(np.mean(scores)))
return np.mean(split_scores), np.std(split_scores)
if __name__ == '__main__':
# class IgnoreLabelDataset(torch.utils.data.Dataset):
# def __init__(self, orig):
# self.orig = orig
#
# def __getitem__(self, index):
# return self.orig[index][0]
#
# def __len__(self):
# return len(self.orig)
image = ImageFolder()
print("Calculating Inception Score...")
#print (inception_score(IgnoreLabelDataset(image), cuda=True, batch_size=32, resize=True, splits=10))
print(
inception_score(image,
cuda=True,
batch_size=32,
resize=True,
splits=10))
convert_to_rgb(),
transforms.ToTensor(),
transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])
]),
'art_Val' : transforms.Compose([
transforms.Resize((256, 256)),
convert_to_rgb(),
transforms.ToTensor(),
transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])
])
}
batch_size = 8
artLen = 300
data_dir = "../../../SwimData/GeoCodes/classification2"
image_datasets = {x: ImageFolder(os.path.join(data_dir, x.split("_")[0]),
data_transforms[x])
for x in ['art_Train', 'art_Val', 'real_Train', 'real_Val']}
realMaster = image_datasets['real_Train']
dataset_sizes = {x: len(image_datasets[x]) for x in ['art_Train', 'art_Val',
'real_Train', 'real_Val']}
class_names = image_datasets['art_Train'].classes
print('Types of classes:',class_names)
classCounts = {x:[image_datasets[x].targets.count(Class) for
Class in np.arange(len(class_names))]
for x in ['art_Train', 'art_Val', 'real_Train', 'real_Val']}
sampleWeights = {x: np.array([1.0/np.array(classCounts[x])[t] for t in image_datasets[x].targets])
for x in ['art_Train', 'art_Val', 'real_Train', 'real_Val']}
import os
from imagefolder import ImageFolder
if not os.path.exists('images_dataset_tar'):
os.makedirs('images_dataset_tar')
im = ImageFolder([
'./images_dataset_ds/train/011k07', './images_dataset_ds/train/015x4r',
'./images_dataset_ds/train/01bqk0', './images_dataset_ds/train/01jfm_',
'./images_dataset_ds/train/01s105'
])
im.to_tar('./images_dataset_tar/train_images_1.tar.gz')
im = ImageFolder([
'./images_dataset_ds/val/011k07', './images_dataset_ds/val/015x4r',
'./images_dataset_ds/val/01bqk0', './images_dataset_ds/val/01jfm_',
'./images_dataset_ds/val/01s105'
])
im.to_tar('./images_dataset_tar/val_images_1.tar.gz')
im = ImageFolder([
'./images_dataset_ds/train/021sj1', './images_dataset_ds/train/02d9qx',
'./images_dataset_ds/train/02s195', './images_dataset_ds/train/034c16',
'./images_dataset_ds/train/03l9g'
])
im.to_tar('./images_dataset_tar/train_images_2.tar.gz')
im = ImageFolder([
transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])
]),
'artVal':
transforms.Compose([
transforms.Resize((256, 256)),
convert_to_rgb(),
transforms.ToTensor(),
transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])
])
}
batch_size = 8
artLen = 1500
data_dir = "../../../SwimData/GeoCodes/classifier4"
image_datasets = {
x: ImageFolder(os.path.join(data_dir, x), data_transforms[x])
for x in ['realTrain', 'realVal', "artTrain", "artVal"]
}
dataset_sizes = {
x: len(image_datasets[x])
for x in ['realTrain', 'realVal', "artTrain", "artVal"]
}
class_names = image_datasets['realTrain'].classes
print('Types of classes:', class_names)
classCounts = {
x: [
image_datasets[x].targets.count(Class)
for Class in np.arange(len(class_names))
]