We covered the preparation for using a pre-existing service-based deep learning model for the task at hand, to predict the species of plant present in a picture. We will be training an image classifier model on samples from five different species of flowers. The model will then try to determine the species to which any image of a flower might belong. However, such models are usually trained on a generally available dataset, and would not have the specificity that might be required at times—for example, in a scientific laboratory. Hence, you must learn how to build your own models for predicting the plant species.
This can be achieved either by training a model completely from scratch or, alternatively, by extending a previously existing model. The upside to training a model completely from scratch is that you have complete control over...