Using the built-in CV algorithms

In this section, we're going to train and deploy models with all three algorithms using public image datasets. We will cover both training from scratch and transfer learning.

Training an image classification model

In this first example, let's use the image classification algorithm to build a model classifying the Dogs vs. Cats dataset that we prepared in a previous section. We'll first train using image format, and then using RecordIO format.

Training in image format

We will begin training using the following steps:

1. In a Jupyter notebook, we define the appropriate data paths:

   import sagemaker
   sess = sagemaker.Session()
   bucket = sess.default_bucket()
   prefix = 'dogscats-images'
   s3_train_path = 
     's3://{}/{}/input/train/'.format(bucket, prefix)
   s3_val_path = 
     's3://{}/{}/input/val/'.format(bucket, prefix)
   s3_train_lst_path = 
     's3://{}/{}/input/train_lst...