number-recognition-app

An interactable tkinter written number recognizer, using Hukyl/mlgo as shared library.

For training the neural network, refer to Hukyl/mlgo.

Preparation

1. Install Go 1.21+ and Python 3.10+.

2. Install Go and Python dependencies.

3. Install Ghostscript.

4. Train a neural network using mlgo, or use a pretrained model dump (see pretrained_models/).

5. From Linux, build the .so file for desired platform (if building from Windows, see section Building).

$ ./compile.sh <build-option>

6. Create .env file in gui/ folder.

SO_PATH="/path/to/built/from/go"
NN_DUMP_PATH="/path/to/model/dump.json"
GS_PATH="/path/to/ghostscript/executable"

Running

To run, just enter the following command:

> python start.py

Building

In case building for Windows from Linux, install mingw:

$ sudo apt-get update
$ sudo apt-get install gcc-mingw-w64  // for 64-bit architecture

When building from Windows for Windows, you can just run the following command:

> go build -o predict.so -buildmode=c-shared predict.go