Running a trained model on consecutive images


So far when I’ve run through the Fastbook examples, it’s applying a trained model to a single image to give the probability of whether it recognises something is present.

Suppose I want to run consecutive images e.g. a series of aerial photographs. Is this something that Deep Learning can be applied to?

Presumably I would be left with a series of probabilistic outcomes for each photograph?

If I have lots of consecutive images for a region then how would I go about explaining/visualising the data for each location?

Grateful if somebody would be willing to clarify as it is something that I am thinking through.


Learner.predict - get prediction for a single item
Learner.get_preds - get predictions for batch of items

This is not clear to me. More information would help.

If you mean you want to show your model an image and predict the region, then you just label your images based on region and it will be a normal multi-class supervised learning problem.