Source private · NDA 2026

VETKA_DWG — photo of a survey sketch → DXF

Computer-vision pipeline that turns a photo of a hand-drawn survey sketch into a topographic DXF. Marker detection, OCR, block library. Cuts manual drafting from 4 hours to 15 minutes.

VETKA_DWG block library audit console — Block-library audit: 444 / 444 blocks pass all five fidelity axes.

Problem

A surveyor draws a field abris by hand, then spends hours transferring numbered points and symbols into CAD. The process is slow, repetitive and error-prone.

Approach

Marker and text detection on a phone photo, followed by OCR for point numbers and labels.
A 444-block GUGK sign library with a fidelity audit against the reference template.
DXF generation with a PySide6/QML desktop shell and CI quality gates.

Results

manual drafting time: 4 h → 15 min
surveying companies in production: 3
GUGK blocks in the library: 444

Stack

Python
OpenCV
PaddleOCR
ezdxf
PySide6 / QML
CI quality gates

What is public

Source private. Commercial use and government clients under NDA; architecture, screenshots and measured results are shown here.