Training a image classification model

I am training an image classification model on google colab. I have successfully trained quite a small dataset however I am wondering if the images I am using are too large, the typical jpeg file size from my camera is around 10MB. There must be a trade off due to file size and colab’s processing speed… would it be better to lower the quality before I annotate the images and train the model? I am using the model to have my cctv recognise squirrels in my garden