utils/transforms.py

import albumentations as A

from albumentations.pytorch import ToTensorV2
from torchvision import transforms as transforms

# Define the training tranforms
def get_train_aug():
    return A.Compose([
        A.MotionBlur(blur_limit=3, p=0.5),
        A.Blur(blur_limit=3, p=0.5),
        A.RandomBrightnessContrast(
            brightness_limit=0.2, p=0.5
        ),
        A.ColorJitter(p=0.5),
        # A.Rotate(limit=10, p=0.2),
        A.RandomGamma(p=0.2),
        A.RandomFog(p=0.2),
        # A.RandomSunFlare(p=0.1),
        # `RandomScale` for multi-res training,
        # `scale_factor` should not be too high, else may result in 
        # negative convolutional dimensions.
        # A.RandomScale(scale_limit=0.15, p=0.1),
        # A.Normalize(
        #     (0.485, 0.456, 0.406),
        #     (0.229, 0.224, 0.225)
        # ),
        ToTensorV2(p=1.0),
    ], bbox_params={
        'format': 'pascal_voc',
        'label_fields': ['labels']
    })

def get_train_transform():
    return A.Compose([
        # A.Normalize(
        #     (0.485, 0.456, 0.406),
        #     (0.229, 0.224, 0.225)
        # ),
        ToTensorV2(p=1.0),
    ], bbox_params={
        'format': 'pascal_voc',
        'label_fields': ['labels']
    })

# Define the validation transforms
def get_valid_transform():
    return A.Compose([
        # A.Normalize(
        #     (0.485, 0.456, 0.406),
        #     (0.229, 0.224, 0.225)
        # ),
        ToTensorV2(p=1.0),
    ], bbox_params={
        'format': 'pascal_voc', 
        'label_fields': ['labels']
    })

def infer_transforms(image):
    # Define the torchvision image transforms.
    transform = transforms.Compose([
        transforms.ToPILImage(),
        transforms.ToTensor(),
    ])
    return transform(image)