Module instance_segmentation 

Instance segmentation algorithms

This module provides instance-level segmentation functionality including:

• Watershed-based instance separation
• Mask IoU and Non-Maximum Suppression
• Panoptic Quality metric
• Instance overlap utilities

Structs

InstanceMask

A single instance produced by an instance segmentation model.

Functions

instance_overlap

Check whether two instance masks overlap (share at least one foreground pixel).

label_map_to_instances

Convert a dense label map (e.g. watershed output) to a vector of InstanceMasks.

mask_iou

Compute intersection-over-union between two binary masks.

mask_nms

Non-maximum suppression on instance masks using mask IoU.

panoptic_quality

Compute Panoptic Quality (PQ), Segmentation Quality (SQ), and Recognition Quality (RQ) for a single semantic class.

watershed_instance

Marker-controlled watershed segmentation.