create a process for tensorflow model training and classification

[cindyloo]

we can use tensorflow, keras layers and a pre-trained model (like ResNet or MobileNet) to classify story images. We can adapt pre-existing models or create our own - it depends on the type of images we care to analyze. As ImageNet is trained on millions of images, a good first analysis would be to instantiate the model, grab feature extractions, and train a top level layer on our own dataset. This dataset would be a subset of topic images that we would then use to make predictions on the rest of the images within that topic

Generally the pseudo code would look something like this:

from tensorflow.keras import layers
from tensorflow import keras
import tensorflow as tf

#gather a dataset from either our own downloaded topic images or a tensorflow dataset
# reshape the images to a numPy array as tensors w/ standard size, likely 28 x 28
# load a pre-trained ResNet model or our own "keras.h5" 
#re-train the model on our dataset
#use that model to evaluate another dataset according to our interests
#save the updated model for next use
#optionally store those predictions