Learn how to train machine learning models for classification and prediction by following the steps ininteractive notebooks. These tutorials integrate Dataflow intoend-to-end machine learning workflows. You can also view the tutorials inGitHub.

Land cover image segmentation

This land classification model uses aTensorFlow framework and satellite data fromGoogle Earth Engine to demonstrate semantic segmentation.The tutorial usesTensorFlow in Vertex AIto train the model, TensorFlow inCloud Run tomake real-time predictions, and Dataflow to make batch predictions.View the code on GitHub.

Weather forecasting time series regression

This weather forecasting model uses aPyTorchframework and satellite data fromGoogle Earth Engine toforecast precipitation for the next two and six hours.The tutorial uses PyTorch to create a fully convolutional network,Vertex AI to train themodel, Dataflow to create the dataset, and PyTorch to make local predictions.View the code on GitHub.

Global fishing watch time series classification

This classification model uses aTensorFlow framework and Maritime Mobile Service Identity(MMSI) location data to classify whether a ship is fishing every hour.The tutorial usesKeras and TensorFlow to train themodel, Dataflow to create the dataset, and Keras inCloud Run to make local predictions.View the code on GitHub.

Wildlife image classification

This classification model uses anAutoML framework tocreate a model trained to recognize animal species from camera trap pictures.The tutorial usesAutoML in Vertex AIto train the model, Dataflow to create the dataset, andVertex AI to make predictions.View the code on GitHub.