Segment Image (Ultralytics YOLO)

How it fits

Typical backends

Caveats

• I/O contractThe output arm is a single downstream-resolved choice fixed at deploy time; every arm reprojects the same instances, so masks, boxes and classes stay aligned.
• Hard constraintA single Ultralytics segmentation weights file is required; a classification, detection-only or pose checkpoint loads but produces empty masks, and the component fails if no usable file is bound.
• Accuracyimg_size should be close to the checkpoint's training resolution; a very different value silently degrades quality.
• Parameter interactionobject_classes empty keeps every class while a non-empty list restricts detection to those ids, and class_thresholds overrides the confidence for specific classes with the rest falling back to threshold.
• Parameter interactioniou sets the overlap bound for non-maximum suppression, trading duplicate suppression against merging nearby distinct instances.
• State lifetimeThe model and all configuration are captured once at startup; changing any setting requires a redeploy.