Hi,I want to processing images and masks together,however,the size of different pictures in my dataset is different.Therefore, I can't do that as X, Y --> [batch_size, row, col, channel].The following code is the code I used before modification.
def augment(x, hflip=True, rot=True):
hflip = hflip and random.random() < 0.5
vflip = rot and random.random() < 0.5
rot90 = rot and random.random() < 0.5
def _augment(img):
if hflip:
img = flip_axis(img, axis=1)
if vflip:
img = flip_axis(img, axis=0)
if rot90:
img = rotation(img, rg=90)
return img
return _augment(x)
def crop_sub_imgs_fn(x, is_random=True):
x = crop(x, wrg=192, hrg=192, is_random=is_random)
x = crop(x, wrg=72, hrg=72, is_random=is_random)
x = augment(x)
return x
train_hr_imgs is a tuple(16,) ---I set the batch-size=16
namely,It contains 16 pictures of different sizes，and the output is a tuple (16,72,72,3)
And now, I have another mask image, called train_mask_imgs.It is also a tuple(16,) like the train_hr_imgs, and their pictures are one-to-one pair.
If I used the same code like
－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－
"mask_4 = tl.prepro.threading_data(train_mask_imgs,fn=utils.crop_sub_imgs_fn, is_random=True)"
－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－
It can also get the tuple(16,72,72,3), but due to random, the results of two (16,72,72,3) is not one to one pair．
Ｉ ｄｏｎ＇ｔ ｋｎｏｗ ｈｏｗ ｔｏ ｓｏｌｖｅ ｉｔ，Ｔｈａｎｋｓ ｆｏｒ ｙｏｕｒ ｈｅｌｐ！
zsdonghao
commented
4 years ago
Hi,I want to processing images and masks together,however,the size of different pictures in my dataset is different.Therefore, I can't do that as X, Y --> [batch_size, row, col, channel].The following code is the code I used before modification.
hr_4 = tl.prepro.threading_data(train_hr_imgs,fn=utils.crop_sub_imgs_fn, is_random=True)
def augment(x, hflip=True, rot=True): hflip = hflip and random.random() < 0.5 vflip = rot and random.random() < 0.5 rot90 = rot and random.random() < 0.5
def crop_sub_imgs_fn(x, is_random=True):
x = crop(x, wrg=192, hrg=192, is_random=is_random)
train_hr_imgs is a tuple(16,) ---I set the batch-size=16 namely,It contains 16 pictures of different sizes，and the output is a tuple (16,72,72,3) And now, I have another mask image, called train_mask_imgs.It is also a tuple(16,) like the train_hr_imgs, and their pictures are one-to-one pair. If I used the same code like －－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－ "mask_4 = tl.prepro.threading_data(train_mask_imgs,fn=utils.crop_sub_imgs_fn, is_random=True)" －－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－－ It can also get the tuple(16,72,72,3), but due to random, the results of two (16,72,72,3) is not one to one pair． Ｉ ｄｏｎ＇ｔ ｋｎｏｗ ｈｏｗ ｔｏ ｓｏｌｖｅ ｉｔ，Ｔｈａｎｋｓ ｆｏｒ ｙｏｕｒ ｈｅｌｐ！