Introducing Amazon Lookout for Vision
As we discussed in Chapter 3, building a custom ML model from scratch is an overwhelming undertaking that requires tremendous time, not to mention that it requires thousands of images along with expert ML practitioners.
With Amazon Lookout for Vision, you don’t need to invest your time and resources in building ML or deep learning pipelines. Instead, you just need to provide normal and anomalous images and the service does the rest. If your images are already labeled, you can directly import them; otherwise, you can use the service’s built-in labeling interface or utilize SageMaker GroundTruth for labeling. Once you have labeled data, you just need to initiate single-click training. Once the model training completes, you will receive results showing the model’s performance.
You can get started with Lookout for Vision with few images (as few as 30 images) but you may want to provide a larger dataset for complex use cases...