---

name: line-segment-fusion description: "Use weighted_boxes_fusion_1d for normalized 1D spans, token-span ensembling, and NLP/NER postprocessing." disable-model-invocation: true metadata: disco-role: operating license: MIT

Line Segment Fusion

Use this sub-skill when the task is to ensemble normalized 1D intervals with weighted_boxes_fusion_1d.

Use this for

• Merging token spans or predictionstring intervals from several NLP models.
• Converting string class labels to numeric labels before fusion, then mapping fused labels back.
• Tuning iou_thr, skip_box_thr, model weights, conf_type, and allows_overflow for 1D span ensembling.
• Working with normalized [x1, x2] spans where x1 <= x2 and the coordinates live in [0, 1].

Do not use this for

• 2D box ensembling.
• 3D cuboid fusion.
• Full benchmark reproduction as the default path.

Start here

• API reference
• Workflow recipes
• Troubleshooting
• Smoke script