Predict_What_You_Draw

• 中文版 說明

This is a hand-drawn image recognition application based on the Tkinter and PIL libraries. Users can draw digits or letters within the application and make predictions using a pre-trained MNIST model or a custom-trained model.

Install

git clone https://github.com/Luffy1225/Predict_What_You_Draw_TensorFLow.git
cd Predict_What_You_Draw_TensorFLow
pip install -r requirements.txt

Run

Important

Before running, ensure that the images folder exists (otherwise, the corresponding training set cannot be found). You can extract the provided images_DATE archive and rename it to images.

Start Predict What You Draw

python main.py

Start AI Training

python AIModel.py

Start FolderManager

python FolderManager.py

Feature of Predict What You Draw

1. Drawing Canvas: Users can freely draw digits on the canvas provided by the application.

2. Model Selection: Multiple pre-trained models are available for users to choose from and switch between for predictions.

Note

It is recommended to use the mnist_model_Epoch_100.keras model.

3. Prediction Result Display: The application displays the predicted digit below the canvas.

4. Clear Canvas: Users can clear all content on the canvas to redraw digits.

5. Save Image (Right click): Drawn images can be saved as part of the training set. You can also using this feature for making your own dataset.

How to Train Your Own Model?

• tutorial

Structure of images Folder

images/
├── ignore/                 # Ignored data, not used for training or testing
│   ├── label_1/            # Images for label 1
│   │   ├── 0.jpg           # Image file
│   │   ├── 1.jpg
│   │   └── ...             # More images
│   ├── label_2/            # Images for label 2
│   │   ├── 0.jpg
│   │   ├── 1.jpg
│   │   └── ...
│   └── label_x/            # Images for other labels
│       ├── 0.jpg
│       ├── 1.jpg
│       └── ...
├── test/                   # Test data, used for validating the model
│   ├── label_1/            # Test images for label 1
│   │   ├── 0.jpg
│   │   ├── 1.jpg
│   │   └── ...
│   ├── label_2/            # Test images for label 2
│   │   ├── 0.jpg
│   │   ├── 1.jpg
│   │   └── ...
│   └── label_x/            # Test images for other labels
│       ├── 0.jpg
│       ├── 1.jpg
│       └── ...
├── train/                  # Training data, used for training the model
│   ├── label_1/            # Training images for label 1
│   │   ├── 0.jpg
│   │   ├── 1.jpg
│   │   └── ...
│   ├── label_2/            # Training images for label 2
│   │   ├── 0.jpg
│   │   ├── 1.jpg
│   │   └── ...
│   └── label_x/            # Training images for other labels
│       ├── 0.jpg
│       ├── 1.jpg
│       └── ...