What was read
Counts are of items the extraction emitted, not of items
present on the sheet. The gap between the two is what the overlay is for.
Could not be placed
openings outside their own wall
0 of 0
openings with no schedule mark 0 of 0
openings with no position or width
0 of 0
named rooms with no wall cycle
0 of 0
schedule rows on this sheet 0
An opening with no mark cannot be joined to the door and window schedule, so its size, type and head height stay unknown however good the geometry is. A room name with no wall cycle has no coordinate in this schema, so it cannot be drawn on the sheet at all — the name was read, the position was not. The flag-dot layer starts off: 619 markers at fit zoom hide the drawing rather than direct the eye. The counts above carry the same information without the noise.
What the colours claim Openings and site regions carry their own reasons, listed above — an opening outside its wall or without a class is a defect in the extraction, not a low score.
Reader confidence — a run-length bucket, not a placement score
The reader's own score for every wall it emitted. Walls
left of the 0.70 line are drawn as SHORT OR FRAGMENTED
RUNS on the sheet — the runs most likely to be a jamb return rather than a
wall. That is what this colour claims, and it is all it claims.
no wall carries a confidence value
Site take-off regions 327 (141 closed, 186 open outlines)
planting 243 of 292 with a species
take-off measured not-the-design-zone
printed by M Hudson 5604 sf
reader union, clipped 5380.1 sf (x0.96)
our closed-area sum 21937.7 sf
a naive sum double-counts nested and overlapping regions, so it reads HIGH by construction; the reader's clipped union is the comparable figure
these areas are NOT the site not-the-design-zone
NONE of the 292 planting symbols fall inside the chosen ring, so it does not enclose the design. On the Davis master plan this ring is the Lot Coverage Graphic's own key diagram, and the figure reproduces MHD's coverage shading against MHD's printed table on the SAME sheet by the SAME author — an internal-consistency check, NOT an independent measurement of the design's hardscape. Do NOT quote `ratio_to_printed` as extraction accuracy against ground truth, and do NOT clip design geometry to `lot_area_sqft`. The ratio above must not be quoted as accuracy.
repeated site copies 2 x 18747 sf, 300.0 ft apart
The same lot is drawn more than once in this modelspace. Everything is shown as found; a consumer that unions or extrudes without picking one instance gets the site more than once.
Rendered coverage
drawn / present
619 / 619
(100.0%)
site regions 327 / 327
planting 292 / 292
Interior is out of scope from 2026-08-09, so `spaces` has
left the denominator. Any coverage change from that is scope-driven, not
an improvement.
Scale
factor 0.08333333333333333 ft per dwg-unit
samples surviving outlier rejection
not reported by this reader
residual
15.0
basis dwg-native
samples 7
dim text sheet len implied 3'-0" GATE 2.250000004901153 1.33333333 3'-0" GATE 3.599999817939633 0.83333338 4'-0" 4.799999757253703 0.83333338 2" 0.1699050638960173 0.98094192 14" 226944.2945525409 5.14e-06 27" 17870248.9888099 1.3e-07 3" 1985279.757525739 1.3e-07
Registration
transform synthetic stand-in — fitted to the plan extent
px per ft 0.9380
plan on sheet
no walls to place
Unresolved — from the reader (5)
converter output repaired R1.0 (dwg2dxf -> dxf_sanitize) 6 desynchronising embedded newlines rejoined; 9 of 127815 entities DROPPED for uninitialised-memory field values (denormal or >1e12 planar coordinates). Dropped, not repaired — a clamped value would have placed a real-looking feature at a fabricated location. Detail in R1.0.sanitize.json hardscape area R1.0: 119 open polylines polyline on a hardscape layer is not closed, so it bounds no area; counted as a region but area left null rather than force-closed region area precision R1.0: 18 arc regions areas include exact circular-segment corrections for polyline bulges; HATCH elliptical edges are approximated by their centre point only planting species R1.0: 36 inserts block insert on a planting layer carries no attribute and no botanical block name; species left null rather than guessed walls/spaces/schedule R1.0 intentionally empty — this is a landscape sheet, not a building sheet; site geometry is in the additive `site` block Assumed — added downstream (6)
open regions emitted as curves, not areas 186 of 327 site regions the drawing outlines these rather than closing them; forcing them shut to make a solid would invent an area the source does not state planting canopy radius (assumed) 279 of 292 symbols canopy_radius_ft is null; markers use the stated defaults {'tree': 8.0, 'shrub': 2.0, None: 2.0} ft by kind. These are position markers, not modelled plants area taken from the reader, not re-derived 18 region(s) with arc bulges area_sqft already carries the exact circular-segment correction; re-deriving from vertices as straight chords would disagree on every curved edge planting with no species 49 of 292 symbols no botanical name resolved from the block or the Land F/X schedule; the position is real, the identification is not the modelspace contains repeated copies of the site 2 boundaries of 18747 sf on L-RSTR-PLINE, centroids up to 300.0 ft apart (b5, b19) separate plan views laid out side by side in one modelspace. Everything is emitted as found; a consumer that unions or extrudes without choosing one instance gets the site more than once no lot boundary used site.takeoff the reader reports measures='not-the-design-zone', not 'design': the ring its take-off chose does not belong to the design. Nothing is clipped, and ratio_to_printed must not be quoted as accuracy
