RandomApply in PyTorch

Buy Me a Coffee☕

*Memos:

• My post explains Compose().
• My post explains RandomInvert().
• My post explains CenterCrop().
• My post explains Pad().
• My post explains OxfordIIITPet().

RandomApply() can randomly apply zero or more transformations to an image with a given probability as shown below:

*Memos:

• The 1st argument for initialization is transforms(Required-Type:tuple/list/torch.nn.Module(transform)): *Memos:
• The transforms are applied from the 1st index in order.
• It must be at least one transformation except when p=0.
• The 2nd argument for initialization is p(Optional-Default:0.5-Type:int or float): *Memos:
  • It's the probability of whether an image is posterized or not.
  • It must be 0 <= x <= 1.
• The 1st argument is img(Required-Type:PIL Image or tensor). *Don't use img=.
• v2 is recommended to use according to V1 or V2? Which one should I use?.

from torchvision.datasets import OxfordIIITPet
from torchvision.transforms.v2 import RandomApply
from torchvision.transforms.v2 import RandomInvert
from torchvision.transforms.v2 import RandomVerticalFlip
from torchvision.transforms.v2 import CenterCrop 
from torchvision.transforms.v2 import Pad

rp = RandomApply(transforms=[RandomInvert(p=1), RandomVerticalFlip(p=1),
                             CenterCrop(size=200), Pad(padding=20)])
rp = RandomApply(transforms=[RandomInvert(p=1), RandomVerticalFlip(p=1),
                             CenterCrop(size=200), Pad(padding=20)], p=0.5)

rp
# RandomApply(RandomInvert(p=1)
#             RandomVerticalFlip(p=1)
#             CenterCrop(size=(200, 200))
#             Pad(padding=20, fill=0, padding_mode=constant))

rp.transforms
# [RandomInvert(p=1),
#  RandomVerticalFlip(p=1),
#  CenterCrop(size=(200, 200)),
#  Pad(padding=20, fill=0, padding_mode=constant)]

rp.p
# 0.5

origin_data = OxfordIIITPet(
    root="data",
    transform=None
)

# `ri` is RandomInvert() and `rv` is RandomVerticalFlip().
# `cc` is CenterCrop() and `pad` is Pad().
ri_rv_cc_pad_p1_data = OxfordIIITPet(
    root="data",
    transform=RandomApply(transforms=[RandomInvert(p=1),
                                      RandomVerticalFlip(p=1),
                                      CenterCrop(size=200),
                                      Pad(padding=20)], p=1)
    # transform=RandomApply(transforms=ModuleList(
    #     [RandomInvert(p=1),
    #      RandomVerticalFlip(p=1),
    #      CenterCrop(size=200),
    #      Pad(padding=20)]), p=1)
)

ri_rv_pad_cc_p1_data = OxfordIIITPet(
    root="data",
    transform=RandomApply(transforms=[RandomInvert(p=1),
                                      RandomVerticalFlip(p=1),
                                      Pad(padding=20),
                                      CenterCrop(size=200)], p=1)
)

ri_rv_cc_pad_p0_data = OxfordIIITPet(
    root="data",
    transform=RandomApply(transforms=[RandomInvert(p=1),
                                      RandomVerticalFlip(p=1), 
                                      CenterCrop(size=200), 
                                      Pad(padding=20)], p=0)
)

ri_rv_cc_pad_p05_data = OxfordIIITPet(
    root="data",
    transform=RandomApply(transforms=[RandomInvert(p=1),
                                      RandomVerticalFlip(p=1),
                                      CenterCrop(size=200),
                                      Pad(padding=20)], p=0.5)
)

import matplotlib.pyplot as plt

def show_images1(data, main_title=None):
    plt.figure(figsize=[10, 5])
    plt.suptitle(t=main_title, y=0.8, fontsize=14)
    for i, (im, _) in zip(range(1, 6), data):
        plt.subplot(1, 5, i)
        plt.imshow(X=im)
        plt.xticks(ticks=[])
        plt.yticks(ticks=[])
    plt.tight_layout()
    plt.show()

show_images1(data=origin_data, main_title="origin_data")
print()
show_images1(data=ri_rv_cc_pad_p1_data, main_title="ri_rv_cc_pad_p1_data")
show_images1(data=ri_rv_pad_cc_p1_data, main_title="ri_rv_pad_cc_p1_data")
print()
show_images1(data=ri_rv_cc_pad_p0_data, main_title="ri_rv_cc_pad_p0_data")
show_images1(data=ri_rv_cc_pad_p0_data, main_title="ri_rv_cc_pad_p0_data")
show_images1(data=ri_rv_cc_pad_p0_data, main_title="ri_rv_cc_pad_p0_data")
print()
show_images1(data=ri_rv_cc_pad_p05_data, main_title="ri_rv_cc_pad_p05_data")
show_images1(data=ri_rv_cc_pad_p05_data, main_title="ri_rv_cc_pad_p05_data")
show_images1(data=ri_rv_cc_pad_p05_data, main_title="ri_rv_cc_pad_p05_data")
print()
show_images1(data=ri_rv_cc_pad_p1_data, main_title="ri_rv_cc_pad_p1_data")
show_images1(data=ri_rv_cc_pad_p1_data, main_title="ri_rv_cc_pad_p1_data")
show_images1(data=ri_rv_cc_pad_p1_data, main_title="ri_rv_cc_pad_p1_data")

# ↓ ↓ ↓ ↓ ↓ ↓ The code below is identical to the code above. ↓ ↓ ↓ ↓ ↓ ↓
def show_images2(data, main_title=None, t=None, p=0.5):
    plt.figure(figsize=[10, 5])
    plt.suptitle(t=main_title, y=0.8, fontsize=14)
    if main_title != "origin_data":
        for i, (im, _) in zip(range(1, 6), data):
            plt.subplot(1, 5, i)
            rs = RandomApply(transforms=t, p=p)
            plt.imshow(X=rs(im))
            plt.xticks(ticks=[])
            plt.yticks(ticks=[])
    else:
        for i, (im, _) in zip(range(1, 6), data):
            plt.subplot(1, 5, i)
            plt.imshow(X=im)
            plt.xticks(ticks=[])
            plt.yticks(ticks=[])
    plt.tight_layout()
    plt.show()

show_images2(data=origin_data, main_title="origin_data")
print()
show_images2(data=origin_data, main_title="ri_rv_cc_pad_p1_data",
             t=[RandomInvert(p=1), RandomVerticalFlip(p=1),
                CenterCrop(size=200), Pad(padding=20)], p=1)
show_images2(data=origin_data, main_title="ri_rv_pad_cc_p1_data",
             t=[RandomInvert(p=1), RandomVerticalFlip(p=1),
                Pad(padding=20), CenterCrop(size=200)], p=1)
print()
show_images2(data=origin_data, main_title="ri_rv_cc_pad_p0_data",
             t=[RandomInvert(p=1), RandomVerticalFlip(p=0),
                CenterCrop(size=200), Pad(padding=20)], p=0)
show_images2(data=origin_data, main_title="ri_rv_cc_pad_p0_data",
             t=[RandomInvert(p=1), RandomVerticalFlip(p=0),
                CenterCrop(size=200), Pad(padding=20)], p=0)
show_images2(data=origin_data, main_title="ri_rv_cc_pad_p0_data",
             t=[RandomInvert(p=1), RandomVerticalFlip(p=0),
                CenterCrop(size=200), Pad(padding=20)], p=0)
print()
show_images2(data=origin_data, main_title="ri_rv_cc_pad_p05_data",
             t=[RandomInvert(p=1), RandomVerticalFlip(p=1),
                CenterCrop(size=200), Pad(padding=20)], p=0.5)
show_images2(data=origin_data, main_title="ri_rv_cc_pad_p05_data",
             t=[RandomInvert(p=1), RandomVerticalFlip(p=1),
                CenterCrop(size=200), Pad(padding=20)], p=0.5)
show_images2(data=origin_data, main_title="ri_rv_cc_pad_p05_data",
             t=[RandomInvert(p=1), RandomVerticalFlip(p=1),
                CenterCrop(size=200), Pad(padding=20)], p=0.5)
print()
show_images2(data=origin_data, main_title="ri_rv_cc_pad_p1_data",
             t=[RandomInvert(p=1), RandomVerticalFlip(p=1),
                CenterCrop(size=200), Pad(padding=20)], p=1)
show_images2(data=origin_data, main_title="ri_rv_cc_pad_p1_data",
             t=[RandomInvert(p=1), RandomVerticalFlip(p=1),
                CenterCrop(size=200), Pad(padding=20)], p=1)

---

---

---

---