TXG Vision Suite

Entry 01

Hypothesis

Asset-tagging contractors record their work in Autodesk Forma, which exports issue PDFs of several hundred pages, each with photos of an installed QR tag. Nobody can open an 800-page PDF, pull every photo, check each QR code against its asset and file the result by hand. The bet: software can do the reading, checking and filing, and leave people only the exceptions.

Entry 02

Experiment

A desktop app on the field engineer's laptop parses the PDF, extracts the photos, builds one collage per issue and checks every QR code, trying 11 detection methods in turn before an optional AI vision fallback. Reviewed photos are sorted into folders (matching, missing, no access) and the matching ones are pushed into IBM Maximo, the client's asset system. A web dashboard and API hold the master asset register and every result, so progress is visible without opening a file. Maximo has no simple upload path, so the upload was built by replaying its classic interface.

Entry 03

Result

From the team's own check: the progress import was run against the client's real export of 4,785 rows, and 4,751 of 4,757 unique asset codes matched (99.87%). A parcel's master register holds about 12,000 assets. The limit is the photos themselves: only about 50 to 60 percent of single QR photos are readable, so that stage is where the effort goes.

Entry 04

Status

Live, and deployed automatically on every update. One open item: a parcel whose extracted records do not match its master register, a data problem still to be fixed by re-importing the register or re-extracting against the right PDF.