Hand OCR Paint is meant to be useful for testing an OCR Neural Network for handwritten characters.
Sample Networks that I trained:
- Alex Net;
- Mobile Net V3;
- KAN Net equivalent to MLP (trained on MNIST);
- Convolutional KAN Net, LeNet but with KAN as Classifier (trained on MNIST).
If you are testing Sample Networks:
- Use [-1, 1] normal range and image inversion if you are using KAN Net or Convolutional KAN Net;
- Use [0, 1] normal range and no image inversion for the others;
Change these parameters according to your training data if you are testing your own network.
You can choose between ONNX and Tensorflow Lite for inference.
- CxImage for image encoding/decoding;
- ONNX neural network engine with DirectML.
- Tensorflow Lite neural network engine.
- Visual Studio 2022 with MFC