# Computer Vision > Computer vision is AI that interprets images and video. In construction, it can read drawings for takeoff, identify symbols and rooms on plans, compare jobsite photos with progress plans, and check PPE compliance in camera feeds. - Stages: Estimating, Operations - Concepts: Computer Vision - Published: 2026-08-28 - Canonical: https://buildr.com/library/computer-vision ## Why it matters in construction Construction runs on drawings and on what the site looks like today. Both are visual. Until recently, turning either into numbers meant a person looking at a screen and counting: doors on a plan, fixtures in a bathroom group, how many floors of curtain wall are hung this week. Computer vision can automate parts of that visual review. In precon, it is most often used for takeoff and plan review. In operations, it can support progress tracking and safety monitoring. When drawings change, a team can rerun the count instead of starting over. ## How it works Computer vision models learn from labeled images. Construction labels might include "door," "receptacle," "column," "hard hat," or "installed drywall." The trained model then identifies those items in new images. Common tasks on drawings: 1. **Classification.** Is this sheet architectural, structural, or MEP? Is it a plan, a section, or a schedule? 2. **Detection.** Where are the doors, windows, and fixtures on this floor plan, and how many of each? 3. **Segmentation.** Which pixels are the building footprint, which are the parking area, which are the room labeled "OR-3"? 4. **Measurement.** Given the scale, how long is this wall and what is the area of this slab? On the jobsite, the same techniques can compare current conditions with the schedule, flag possible missing fall protection, or count deliveries at the gate. The output is usually a list of detected objects with their positions and confidence scores. Useful tools draw those detections on the sheet so an estimator can see what was counted. A total without the marked-up source is hard to verify. ## Example in practice Consider a commercial GC estimating a 3-story, 140,000 sq ft office building with an addendum arriving four days before bid. The addendum revises 22 sheets, mostly partition changes on levels 2 and 3. The estimator runs computer vision takeoff on the revised set. The tool counts 318 doors, up from 302, and 1,140 linear feet of new partition, then remeasures the affected ceiling areas. It marks each detection on the plan and flags nine doors at 60 percent confidence where revision clouds partly cover the tags. The estimator reviews those nine, confirms seven, and rejects two that were cased openings. The team updates the drywall and doors packages for the subs. ## Go deeper - [AI for Estimators: 5 Practical Uses that Work Today (Not Someday)](/blog/ai-for-construction-estimators.md) - [The Future of AI in Construction (Updated for 2026)](/blog/artificial-intelligence-construction.md) ## Related terms - [AI Takeoff](/library/ai-takeoff.md): AI takeoff uses computer vision and language models to read construction drawings, identify building elements, and count or measure quantities so estimators can price a project without tracing every sheet by hand. - [OCR vs. LLM Document Understanding](/library/ocr-vs-llm-document-understanding.md): OCR converts scanned construction documents into text. LLM document understanding reads that text, or the page image itself, and interprets its meaning. Precon tools often use both: OCR handles the pixels and the language model interprets the content. - [AI Plan and Spec Review](/library/ai-plan-and-spec-review.md): AI plan and spec review uses language models to compare a construction drawing set with its specification book. It flags conflicts, missing information, and risk items for the preconstruction team to review before pricing. - [Large Language Model (LLM)](/library/large-language-model.md): A large language model (LLM) is an AI system trained on large amounts of text that can read, summarize, and generate language. In construction, it can help parse RFPs, sub proposals, and specifications for preconstruction teams. ## FAQ ### How is computer vision different from a language model? A language model works on text. Computer vision works on pixels. Newer multimodal models combine both, so one system can look at a floor plan and describe what it sees in words, but the underlying skills are still distinct. ### How accurate is computer vision on construction drawings? On clean, consistently drafted plan sets it can count doors, fixtures, and outlets with high accuracy. Accuracy drops on hand-marked revisions, overlapping layers, and sets that mix drafting standards, which is why takeoff tools show what they counted so an estimator can spot-check. ### Do we need special cameras on site? No. Most jobsite vision tools work from ordinary phone photos, 360 cameras, or existing security feeds. Consistent capture routines matter more than camera quality.