Closed zhaizhai closed 8 years ago
Here are some preliminary test cases I made for stroke recognition. The top-level fields are
character: character data
matches: indices of correct strokes matched so far
stroke: stroke the user made
expected: expected result of matching user stroke
While doing that I realized it might be useful to account for a little bit of context about a stroke. In particular, it seems I had some issues when making two strokes in quick succession. So maybe we should track whether the current stroke happened immediately after the preceding one.
Merged. Thanks for help!
The user's strokes will be collected by the server, and we can use that data to help test/improve stroke recognition.