rtzm / spellcast Goto Github PK

Glyph class should contain the data about the shape being drawn (difference at every point) and also maintain a <canvas> element that is the view layer for that glyph. Clarify what kind of object the OCR handwriting analysis program will expect, and whether it would be better for it to read in from the canvas element or from the Glyph object directly.

Write tests for Glyph's methods that maintain state and keeps up to date with the canvas element. Write tests for the GlyphWriter's write() method that takes in a new vector of change and draws from current point to that new vector.

Running this on my iPhone Chrome v 65 or so doesn’t work

Need to implement a test suite, and then adapt the optflowanalysis.js algorithm into a single function that takes a video and the analysis (LK) parameters as a arguments, and which can be tested.

It’s not very intuitive writing process, because we’re startjng at middle point. In most L to R writing programs your cursor starts top left. Figure out the UX of how this gesture feels more natural.

Use constraints to set ideal frame rate and camera resolution. Downsample resolution when processing based on what the ideal width and height should be (using maximum for what makes the smoothest lines in the algorithm)