RandomScale
Randomly resize the input. Output image size is different from the input image size.
scale_limitscaling factor range. If scale_limit is a single float value, the range will be (-scale_limit, scale_limit). Note that the scale_limit will be biased by 1. If scale_limit is a tuple, like (low, high), sampling will be done from the range (1 + low, 1 + high). Default: (-0.1, 0.1).
interpolationflag that is used to specify the interpolation algorithm. Should be one of: cv2.INTER_NEAREST, cv2.INTER_LINEAR, cv2.INTER_CUBIC, cv2.INTER_AREA, cv2.INTER_LANCZOS4. Default: cv2.INTER_LINEAR.
mask_interpolationflag that is used to specify the interpolation algorithm for mask. Should be one of: cv2.INTER_NEAREST, cv2.INTER_LINEAR, cv2.INTER_CUBIC, cv2.INTER_AREA, cv2.INTER_LANCZOS4. Default: cv2.INTER_NEAREST.
area_for_downscaleControls automatic use of INTER_AREA interpolation for downscaling. Options:
- None: No automatic interpolation selection, always use the specified interpolation method
- "image": Use INTER_AREA when downscaling images, retain specified interpolation for upscaling and masks
- "image_mask": Use INTER_AREA when downscaling both images and masks Default: None.
pprobability of applying the transform. Default: 0.5.
>>> import numpy as np
>>> import albumentations as A
>>> import cv2
>>>
>>> # Create sample data for demonstration
>>> image = np.zeros((100, 100, 3), dtype=np.uint8)
>>> # Add some shapes to visualize scaling effects
>>> cv2.rectangle(image, (25, 25), (75, 75), (255, 0, 0), -1) # Red square
>>> cv2.circle(image, (50, 50), 10, (0, 255, 0), -1) # Green circle
>>>
>>> # Create a mask for segmentation
>>> mask = np.zeros((100, 100), dtype=np.uint8)
>>> mask[25:75, 25:75] = 1 # Mask covering the red square
>>>
>>> # Create bounding boxes and keypoints
>>> bboxes = np.array([[25, 25, 75, 75]]) # Box around the red square
>>> bbox_labels = [1]
>>> keypoints = np.array([[50, 50]]) # Center of circle
>>> keypoint_labels = [0]
>>>
>>> # Apply RandomScale transform with comprehensive parameters
>>> transform = A.Compose([
... A.RandomScale(
... scale_limit=(-0.3, 0.5), # Scale between 0.7x and 1.5x
... interpolation=cv2.INTER_LINEAR,
... mask_interpolation=cv2.INTER_NEAREST,
... area_for_downscale="image", # Use INTER_AREA for image downscaling
... p=1.0 # Always apply
... )
... ], bbox_params=A.BboxParams(coord_format='pascal_voc', label_fields=['bbox_labels']),
... keypoint_params=A.KeypointParams(coord_format='xy', label_fields=['keypoint_labels']))
>>>
>>> # Apply the transform to all targets
>>> result = transform(
... image=image,
... mask=mask,
... bboxes=bboxes,
... bbox_labels=bbox_labels,
... keypoints=keypoints,
... keypoint_labels=keypoint_labels
... )
>>>
>>> # Get the transformed results
>>> scaled_image = result['image'] # Dimensions will be between 70-150 pixels
>>> scaled_mask = result['mask'] # Mask scaled proportionally to image
>>> scaled_bboxes = result['bboxes'] # Bounding boxes adjusted to new dimensions
>>> scaled_bbox_labels = result['bbox_labels'] # Labels remain unchanged
>>> scaled_keypoints = result['keypoints'] # Keypoints adjusted to new dimensions
>>> scaled_keypoint_labels = result['keypoint_labels'] # Labels remain unchanged
>>>
>>> # The image dimensions will vary based on the randomly sampled scale factor
>>> # With scale_limit=(-0.3, 0.5), dimensions could be anywhere from 70% to 150% of original- The output image size is different from the input image size.
- Scale factor is sampled independently per image side (width and height).
- Bounding box coordinates are scaled accordingly.
- Keypoint coordinates are scaled accordingly.
- When area_for_downscale is set, INTER_AREA interpolation will be used automatically for downscaling (scale < 1.0), which provides better quality for size reduction.