Skip to content

Latest commit

 

History

History
23 lines (20 loc) · 1.2 KB

index.md

File metadata and controls

23 lines (20 loc) · 1.2 KB

Modules

Pre-trained TensorFlow Hub modules are already available for a variety of tasks. They are described in the following documents:

  • Image Modules, which lists modules trained extensively to classify objects in images. By reusing their feature detection capabilities, you can create models that recognize your own classes using much less training data and time.
  • Text Modules, which lists modules with pre-trained text embeddings that can be used to classify text. These can be simple lookup tables or more complicated designs, usually accepting full sentences or paragraphs.
  • Other Modules, which lists modules for other types of tasks, such as mapping from latent space to images or extracting deep local features.

As in all of machine learning, fairness is an important consideration. Modules typically leverage large pretrained datasets. When reusing such a dataset, it’s important to be mindful of what data it contains (and whether there are any existing biases there), and how these might impact your downstream experiments.