Utilities

Small helpers shared across the model pages: loading an image or video into the array shape the processors expect, drawing results, and the class-name lists that turn label ids back into words. One page per module in kerasformers/utils/.

Page	Module	Covers
Image utils	image_util.py	load_image, get_data_format, standardize_input_shape
Video utils	video_util.py	load_video, VideoMetadata, sample_frames, the three decoders
Visualization utils	visualization_util.py	plot_detections, plot_segmentation, plot_depth, plot_sam_masks
Class labels	labels_util.py	COCO, ADE20K, Cityscapes, and VOC class-name lists

from kerasformers.utils import (
    load_image, get_data_format, standardize_input_shape,
    load_video, sample_frames, VideoMetadata, VIDEO_DECODERS,
    plot_detections, plot_segmentation, plot_depth, plot_sam_masks,
)
from kerasformers.utils.labels_util import COCO_80_CLASSES, ADE20K_150_CLASSES

Everything here is optional. The processors accept plain PIL images and NumPy arrays, so you can skip these entirely and use PIL, OpenCV, and Matplotlib directly.

See also Main Classes for the model, processor, and generation bases.