Se non è presente alcun dispositivo di cattura video, o questo viene disconnesso, l'applicazione si blocca e non visualizza alcun messaggio di errore.

Some fixes for better reproducibility

1. An installation guide

2. Specify which packets we need + python version + packets version

pip install tensorflow opencv-python pillow pygame pyautogui sklearn tflite scikit-learn tflite-runtime

3. Windows

Note that from python 3.10 tflite-runtime manual build is required.
A possible fix is to use Python 3.9.

4. Model upload

In the release would be great to have tflite model. The correct model directory must be Poke-Pi-Dex-main\app\resources\classifier_model. Remember to check the model filename (see pokemon_classifier_tflite.py).

5. Path fix

best_classes.npy must be moved from pokemon-classifier/src a /app/resources/classifier_model

If using the tflite model converted from PyTorch decomment the following line

img = np.transpose(img, [0, 3, 1, 2])