Skip to content
Nat Lee

Keyboard shortcuts

Turn these off if they get in the way of a screen reader or voice control. The setting is saved in this browser.

1
Home
2
Projects
3
About
4
Experience
/
Search projects
?
Show this panel

Computer Vision

simply-blur-detector

CNN meets Laplacian, in your browser

Drop in a picture and compare a small Keras CNN with Laplacian variance, including a window-by-window blur heat map. No upload needed.

Source1010 stars on GitHub(opens in a new tab)Live demo(opens in a new tab)
Role
Creator & Maintainer
Period
2019 – 2026.11
Status
Completed
Theme
AI, vision & medical imaging
simply-blur-detector, screenshot 1

Overview

A TensorFlow / Keras CNN trained on sharp illustrations and copies blurred with blur-generator. The 2026 version includes original training artwork, x86 and ARM Docker support, and ONNX export for a browser demo that compares the CNN with Laplacian variance. Its verdict is a model estimate, with both methods visible for comparison.

The problem

  • Matching preprocessing between Python training and browser inference

What I did

  • The same equalisation, resize and normalisation steps across training and export

Results

  • Interactive GitHub Pages demo with local image processing
  • Reproducible training assets and model evaluation in the repository

Features

  • Browser-local inference and patch-by-patch heat maps
  • CNN and Laplacian variance side by side
  • Original sharp and blurred training illustrations
  • Docker training and ONNX export
  • AI, vision & medical imaging33 stars on GitHub

    QWOP with PPO

    A PyTorch PPO agent plays the real QWOP browser game through Playwright. Six evaluation races, six 100 m finishes. Grace was not part of the reward function.

    2019Completed
  • AI, vision & medical imaging4040 stars on GitHub

    Brain MRI Pipeline

    An end-to-end Nipype pipeline with FSL FAST, K-means and ANTs Atropos segmentation, tissue volumes, bias-correction metrics and an interactive 3D brain.

    2018Completed
  • AI, vision & medical imaging11 stars on GitHub

    Street Scene OCR

    Compare EasyOCR, RapidOCR and PaddleOCR on Traditional Chinese street signs, with text boxes, timings and a small public-domain benchmark.

    2026Completed