v1.2 - 漢字検定

new Features:

• recognize:

  • all 漢字 from 漢字検定
  • ひらがな (also historical ones: ゑ, etc.)
  • カタカナ (also historical ones: ヱ, etc.)

Changes:

• handle class imbalance better
• moved fonts into root directory

v1.1

This version is an improved version of v1.0 with no additional features.
This model generalizes better and can handle slight variations better.

v1.0

This release contains the tflite and the tf model as well as all characters which can be recognized.
The characters (labels) are in the same order as the output of the CNN.

Features:

• can recognize ~3000 kanji characters