Mask

A box covers the object and the background around it. A segmentation model also gives a mask: the pixels that belong to the object. pf.annotate.mask colors those pixels.

Python
import pixelflow as pf
from mayaku import from_pretrained

predictor = from_pretrained("mayaku-n-seg")
result = predictor("street.jpg")
detections = pf.from_mayaku(result, labels=predictor.class_names)
detections = detections.filter_by_confidence(0.5)

image = pf.read_image("street.jpg")
image = pf.annotate.mask(image, detections)
pf.save_image("street-masks.jpg", image)

See-through colored masks following the shapes of the people and motorcycles in a street photo

Detections without a mask are skipped. Where two masks overlap, the one later in the list takes the pixel.

A mask must have the same width and height as the image; otherwise mask raises a ValueError. That is the case for masks from every converter, as long as you draw on the image at the size you gave the model: resize the picture after drawing, not before.

To draw only the edge of each object, use polygon.

Parameters

Parameter Default
frame RGB image, drawn on in place
detections the Detections whose masks to draw
opacity 0.5 from 0 (invisible) to 1 (solid color)
colors the default palette list of RGB colors, picked by class_id (see colors)