Using a conventional neural network to classify images Pokemon and Digimon.
- Python 3.11.5+
- NumPy 1.24.3+
- scikit-learn 1.3.0+
- tensorflow 2.13.0+
- pandas 2.1.0+
- Clone this repository.
- Open the terminal and navigate to the directory where the repository is located.
- Download the Pokemon vs. Digimon dataset from Kaggle. Unzip the data and copy the folder
automlpoke
to thedata
folder you just created.
This process is not required, since the trained model can be found as models\pokemon_vs_digimon_CNN_latest.h5
. Training the model will create (or overwrite) the file results\pokemon_vs_digimon_CNN_latest.h5
, as well as create a CSV file containing information about the training process.
- Run the following command from the root of your repository:
python main.py train
- Run the following command from the root of your repository:
python main.py classify
- Input the relative or absolute path of the image file you want to classify.
This process will output a file called bulk_classification_results_X.csv
where X is the smallest positive integer that does not cause a naming conflict.
- Create a folder exclusively containing images of Pokemon or a folder exclusively containing images of Digimon.
- Run the following command from the root of your repository:
python main.py classify-bulk
- Input the relative or absolute path of the folder of images you want to classify.
Thank you to Vinícius Vale for the initial training data I used for this model.
All Pokemon and Digimon are trademarks of their respective creators. No images of Pokemon or Digimon are included in this repository.
V. S. Vale, "Pokemon vs Digimon Image Dataset," Kaggle, 2020. [Online]. Available: https://www.kaggle.com/datasets/vsvale/pokemon-vs-digimon-image-dataset.