About AutoScan AI

AutoScan AI is an end-to-end Automatic Number Plate Recognition (ANPR) system. Upload a photo of a vehicle and it finds the vehicle, isolates the number plate, reads the characters and identifies the plate's region and type — then keeps a searchable history of every scan.

Use cases

  • Parking lot entry and exit logging
  • Gated society and campus vehicle records
  • Toll and fleet tracking prototypes
  • Learning project for computer vision pipelines

What I learned

  • Exporting a PyTorch model to ONNX and writing the pre/post-processing (letterbox, NMS) by hand.
  • Combining deep learning with classical computer vision instead of training a model for every sub-task.
  • Measuring and cutting memory use so an ML service fits on free cloud hosting.
  • Building a full product around a model: auth, storage, history, analytics and an API.

Tech stack

Machine learning

YOLOv8nONNX RuntimeOpenCVNumPyGemini Vision

Backend

PythonFastAPISQLAlchemyPostgreSQL (Neon)

Frontend

Next.js 15React 19TypeScriptTailwind CSSNextAuth

Deployment

VercelRenderGitHub
M

Built by

Munna Kumar · CodeWithMunnaX

Developer working on machine learning, data science and full-stack web projects.

GitHub codewithmunnax.com
See how the ML pipeline works →