Approaches Index

Status: Historical and retained-method index

Authority: Accepted decision DL-010, strategy freeze and evaluation reports, current contract

Last reviewed: 2026-08-14

Question

Which earlier approaches matter, and what did Drawing Lab retain from them?

Short answer

Each history page answers four questions. What did we try? Why did the method help? Why did the method fail or stop? What did we retain?

Index

Topic	Page
Whole-sheet raster processing and information islands	Whole Sheet to Information Islands
Boundary detection, magic wand, and RGB deviation	Boundary, Wand, and RGB Deviation
Topology and view candidates	Topology and View Candidates
Outline versus centreline and centreline recovery	Outline Versus Centreline
Independent vector comparison architecture	Vector Comparison Architecture

Also covered elsewhere

Topic	Where
Original table heuristics and external detector	Proposal Versus Verification
Table providers and TableFormer crop-only role	Provider Comparison
Original dimension assembly and Fast Dimension Lines	Fast Dimension Lines
Line/Arc and Vector Lines	Line/Arc and Vector-Line Comparisons
OCR, symbols, arrowheads	OCR, Symbols, and Arrowheads
Ridge, Potrace, Inkscape	Geometry, Lines, Arcs, and Ridge

Open Questions

• Should later history pages add dated evidence cards for every evaluation number?

Sources

• Accepted decision: DL-010
• Current contract: table-dimension-detection-architecture.md (2026-08-12)
• Wiki surface plan required algorithm history table (plan draft)