Steel takeoff from scanned drawings
Scanned sets show up on renovations, public bids and older buildings: raster images, skewed pages, faded member sizes and title-block scales that stopped being true when someone resized the print. You can still take off from them, but only with more checking and a written record of what you assumed.
- Calibrate twice
- Dimensions beat scaling
- OCR has limits
- Record every assumption
Three ways information is lost
- Scale drifts. A set plotted at a different size, or scanned with "fit to page," no longer matches the title-block scale. Half-size sets are common and fine if you know it; unknown resizing is not.
- Text degrades. Member sizes and small dimensions are the first to go. W12x26 and W12x28 look identical on a third-generation copy.
- Geometry distorts. Pages scan slightly rotated or skewed, and some scanners stretch one axis more than the other.
None of these stop a takeoff. Each needs a specific check.
Checks on every sheet
- Find a known dimension. Use a long dimension string or grid-to-grid spacing, the longer the better, since errors in picking endpoints shrink in proportion.
- Calibrate from it, not from the title block. In Bluebeam Revu, use Calibrate in the Measurements panel and click the two ends of the known dimension; other takeoff tools have the same function.
- Check a second dimension in the other direction. If horizontal and vertical disagree by more than a percent or so, the scan is stretched. Calibrate per axis if your tool allows, or rely on written dimensions only.
- Treat enlarged details as separate scale regions. A detail at 3/4" = 1'-0" on the same sheet as a 1/8" plan needs its own calibration.
- Deskew if grids aren't square to the page; measured lengths on a rotated page are still right, but counts by area selection get messy.
When you can't read it
- Cross-reference. A size that's illegible on the plan may be clear in a beam schedule, a section, or the same beam's mark on another floor.
- Use typical framing carefully. Repetitive bays usually repeat sizes. Use that as an assumption, and write it down.
- OCR helps, with limits. Running OCR (in Bluebeam, Document > OCR) makes text searchable, which helps you find every "W16" on a sheet. It does not make an unreadable size readable, and OCR mistakes on small, angled text are common. Treat OCR hits as leads to check visually.
- Ask. Send an RFI listing illegible sizes by grid line and sheet. One RFI with a table beats twelve emails.
- Never guess silently. If you carry an assumed size, the takeoff row says so.
Vector PDFs, CAD or originals
Ask for vector PDFs or the original CAD where they exist; even on older buildings, the architect may have CAD files that were printed and scanned for the bid set. For existing buildings with no drawings at all, you're into field measuring. Price a site visit, or carry an allowance and qualify the bid.
The AISC Code (§3.5) requires design drawings to be clearly legible and drawn to an identified scale. If a bid set isn't legible, saying so in an RFI is reasonable, not a complaint.
Calibration check
Illustrative. A scanned framing plan shows grid A–B dimensioned at 30'-0" and grid 1–2 at 25'-0".
Calibrate on A–B (30'-0")
Measure 1–2 with that calibration: 25'-7"
Difference: 7 in on 25 ft ≈ 2.3% → vertical axis stretched
Action: use written dimensions for all N–S members; calibrate
separately for N–S runs; note "scaled, ±2%" on any unscaled member
A 2% error on beam lengths is small in tonnage, but it compounds on long continuous runs (edge angles, rails, deck edge) and on areas, where a 2% error in each direction is about 4%.
What to write down
Every takeoff from a scan should carry a short log at the top:
- Sheets that were scanned vs. vector.
- Calibration per sheet: known dimension used, check dimension, error found.
- Members with assumed sizes, by grid and sheet.
- RFIs sent and answers received.
- Allowances carried because of legibility.
That log is what protects the number when a change order discussion starts six months later.
Member counts off the drawings, with the sheet attached.
RenderDraw's AI takeoff proposes steel member counts and lengths from plans, framing plans and schedules, links every row to its source, and sends low-confidence items to your estimator before the handoff to pricing.
- Counts and lengths by member. Proposed from the drawings and kept tied to the framing plan or schedule they came from.
- Uncertain items flagged. Items the AI can't confirm go to review instead of being guessed.
- Addenda through the same review. Send the changed sheets through takeoff and review, then reprice the members that changed.
- Your estimator decides. Connections, erection, coatings, risk and margin stay your call.
Test it on a real steel package
Send one structural set, ideally with an addendum. Sample files stay private while RenderDraw assesses the request. No obligation.
RenderDraw does not publish accuracy percentages or time-saved claims. Test it on your own plan set.
Questions estimators ask.
Can you do a steel takeoff from scanned drawings?
Yes, with extra checks. Calibrate each sheet from a known dimension, verify with a second dimension in the other direction, rely on written dimensions over scaled ones, and record every assumed size and RFI.
Why doesn't the title block scale work on a scan?
Prints are often reduced, enlarged or scanned with "fit to page," so the drawing no longer matches the stated scale. Always calibrate against a dimension on the sheet.
Do written dimensions or scaled dimensions govern?
Written dimensions. AISC 303-22 Section 3.4 states that when scaled dimensions and written figures disagree, the figures govern.
Does OCR fix illegible member sizes?
No. OCR makes legible text searchable. It can misread small or degraded text, and it can't recover text that isn't legible. Check OCR results visually.
What should I do if sizes are unreadable?
Cross-reference schedules and other floors, use typical framing only as a documented assumption, and send an RFI listing every illegible size by grid and sheet.
Can RenderDraw read scanned steel drawings?
A legible PDF set is the most reliable start; test your scanned sets in the free sample. Legibility matters: RenderDraw flags low-confidence items for your estimator rather than guessing, and checking scale against known dimensions stays part of review. Get a free sample takeoff.
Flags, not silent guesses.
Send a scanned set. RenderDraw proposes quantities with sources and flags items it could not confirm for estimator review.
- Upload the plans, specs and schedules, plus an addendum if you have one.
- RenderDraw reviews the set and the scope you want tested.
- We follow up with extracted quantities, flagged items and next steps.
Where these numbers come from
We checked this guide against the standards, manufacturer documents and industry references below (last reviewed November 2026). Figures marked as rules of thumb are typical ranges, not requirements; your project documents, local code and your own cost history govern.
- ANSI/AISC 303-22 Code of Standard Practice (free from AISC) — §3.4 discrepancies, §3.5 legibility and commentary on scaling for estimates
- Bluebeam: Set the page scale on your drawing (calibrate)
- Bluebeam: OCR
- RenderDraw: AI takeoff from PDF drawings