Working with a visual search input pipeline using tf.data

The TensorFlow tf.data API is a highly efficient data pipeline that processes data an order of magnitude faster than the Keras data input process. It aggregates data in a distributed filesystem and batch processes it. For further details, refer to: https://www.tensorflow.org/guide/data.

The following screenshot shows an image upload time comparison of tf.data versus the Keras image input process:

Note that 1,000 images take about 1.58 seconds, which is about 90 times faster than the Keras image input process.

Here is some common features for tf.data:

• For this API to work, you need to import the pathlib library.
• tf.data.Dataset.list_files is used to create a dataset of all files matching a pattern.
• tf.strings.splot splits the file path based on a delimiter.
• tf.image.decode_jpeg decodes a JPEG image into a tensor (note...