In the previous module, you have seen the power of tf.data APIs in TensorFlow 2.0 reading complex data types from any storage.

In this module, you will see an example of data ETL from raw images to input into tensors, then apply transfer learning (which is how a lot of future models for end-users will be built at companies) to build an emotion classification model. This use case would be really cool for deployment in the next module, where you can see the inference of your own facial expressions. Note however that at the moment, TF 2.0 models converted to tflite are incompatable with the devices. Follow us to get the latest updates post workshop when the TensorFlow team has fixed the issue.

What you’ll learn

In this lab, you will learn to:

• Examine and understand the data (not exhaustive)
• Build an example input pipeline (there are many ways to build an input pipeline)
• Compose your model:
• Choose a suitable pre-train model
• Choose the sub-model you will depend on to build your model
• Append the suitable classification layers at the end
• Train your model
• Evaluate your model
• Tune and/or update the architecture of your model

#tensorflow #python #deep-learning #machine-learning #data-science