img = cv2.imread(path)
img = cv2.resize(img,(IMG_SIZ,IMG_SIZ),3)
Found,
If you are training a custom model with a number of class != 80,
so to speak something different than the pre-trained weights ;
you will have to create your own class file (class.names) under data
and set this new file in the config.py file as during the model creation,
the number of class is automatically deduces
trainX = train[:, 1:].reshape(train.shape[0],1,28).astype( 'float32' )
X_train = trainX / 255.0