carlz32/draw-guess

Draw and guess

Description

This project is intend to learn Machine Learning. You can follow @Radu on youtube and code step by step.

Knowledge Required

• HTML, CSS and Vanilla JS
• JavaScript Canvas API
• Basic NodeJS API
• Vector 2D Math

TODO

• Instead of redrawing everything on the canvas when emphasis a sample, use a transparency canvas to only redraw the sample.

• Best K value

• Using an overlay canvas to optimise performance when hovering and drawing

• Long list rendering perf

• Calculate the sample's width and height in a more reasonable way.

  • using something called Convex Hull
  • the bound is called Minimum Bounding box
  • the algorithm is Graham Scan Algorithm

• Calculate a probability for the classification to be correct, based on the number of neighbors of the predicted class.

Tips

• JS files in node folder need to be run in the same directory, not the root