← All work

Research & models / 2024—25

Self-Learning Meter OCR.

An offline Android foundation for capturing and reading meter images in low-connectivity settings.

PracticeComputer vision · mobile field work
My partOCR research, model training and Android integration.
Explore00 visual assets
Research & modelsSelf-Learning Meter OCRSystems shaped for people.
01 / Context

The problem behind the work.

Field reading needs a capture path that can work with constrained devices and intermittent connectivity.

02 / Contribution

What I built.

  1. 01

    A mobile capture and reading workflow with local persistence and repeatable OCR training.

  2. 02

    Low-memory inference experiments using PaddleOCR, OpenCV and Android tooling.

03 / Where it stands

Public repositories document the OCR foundation; they are linked as source, not as a live field deployment.

Technology and methods

JavaAndroidPaddleOCROpenCVSQLite
Next project / Research & modelsDocument intelligence