Google Code-in

Most of the Deep Learning based image classifiers need a considerable amount of data to perform well and, getting data is not easy in most of the cases. There are a lot of pre-trained image classifiers available. We can use these pre-trained classifiers and retrain them on a new but related problem. This method is known as transfer learning. It makes training easier when we have a limited number of samples. This is a tutorial on transfer learning using tf.keras .

A trained machine learning model is not of much use unless it can be integrated with other applications. One of the ways to do this is - serving the model as an API. TensorFlow Serving is a part of the TensorFlow ecosystem which can be used to deploy models.

The objectives of this task are

1. Use a pre-trained image classifier and train it on a new dataset (pick any).
2. Export the trained model using saved_model format.
3. Deploy the exported model using tensorflow serving.