Compute IoU Between Masks#

Intersection over Union (IoU) measures how much two masks overlap. It’s the standard metric for comparing segmentations.

Two masks#

from rlemasklib import RLEMask

iou = RLEMask.iou(mask1, mask2)

This returns a float between 0 (no overlap) and 1 (identical).

For comparing many masks against many others:

# Returns a matrix of shape (len(masks_a), len(masks_b))
iou_matrix = RLEMask.iou_matrix(masks_a, masks_b)

With COCO-format dicts:

import rlemasklib

iou = rlemasklib.iou([mask1_dict, mask2_dict])

If you need the intersection and union areas separately:

intersection = mask1 & mask2
union = mask1 | mask2

iou = intersection.area() / union.area() if union.area() > 0 else 0.0