Cross-Platform Ship Draft Reader -- 2

I’m putting together an iOS- and Android-ready app that lets surveyors read a vessel’s drafts with nothing more than the phone camera and a solid bit of computer-vision logic. The idea is simple for the user, but the tech under the hood needs to be rock–solid and totally offline-capable.

How it should work
• A surveyor logs in and chooses a ship from an onboard list. Each entry carries full particulars, the relevant stability-booklet data, and a history of past surveys.
• At the chosen draft mark—forward, midship, or aft on either port or starboard—the user points the camera. The app auto-detects the number, then immediately shows it back so the surveyor can confirm or correct before saving.
• Sea-side readings are handled by lowering the phone (camera down) or by a quick video capture that the app parses frame by frame.
• Once all four required drafts are locked in, the calculation module runs the stability-booklet formulae and presents the completed survey.

Key requirements
• Accurate, low-latency number recognition that remains fully functional when the device is offline.
• Clean UI/UX for review and manual override of each camera reading.
• Local storage of ship data, booklet calculations, and every completed survey for later export or sync.
• Consistent performance on both iOS and Android—feel free to recommend Flutter, React Native, or another cross-platform stack if it meets the brief.
• Source code delivered with clear build instructions and unit tests for the recognition engine and calculation routines.

I’m aiming for reliability equal to shipboard standards, so we’ll set acceptance criteria around detection accuracy and calculation repeatability once you’re onboard. If you have proven experience with OCR/computer-vision modules (Tesseract, Vision OS, ML Kit, etc.) and can package it all into a lightweight offline app, let’s talk timelines and milestones.