What Is the Best Input Pipeline to Train Image Classification Models with tf.keras

 By Medium - 2021-02-04

Description

When we start learning how to build deep neural networks with Keras, the first method we use to input data is simply loading it into NumPy arrays. At some point, especially when working with images…

Summary

  • ?
  • Once the two parts of the dataset are finished, they are combined using the zip method, which is similar to the python function with the same name.
  • example of bad input pipeline, as presented by the profiler Another detail that seemed to make a good difference, especially when working with a TPU, is to parallelize the map method.
  • This is related more to the math behind neural network training than with data pipelines, but I saw an increase of 10 accuracy points just by shuffling the data correctly.

 

Topics

  1. Backend (0.39)
  2. Machine_Learning (0.22)
  3. Database (0.17)

Similar Articles

Datasets should behave like git repositories

By DAGsHub Blog - 2021-01-18

Create, maintain, and contribute to a long-living dataset that will update itself automatically across projects, using git and DVC as versioning systems.

14 Data Science projects to improve your skills

By KDnuggets - 2020-12-01

There's a lot of data out there and so many data science techniques to master or review. Check out these great project ideas from easy to advanced difficulty levels to develop new skills and strengthe ...

30 Most Asked Machine Learning Questions Answered

By Medium - 2021-03-18

Machine Learning is the path to a better and advanced future. A Machine Learning Developer is the most demanding job in 2021 and it is going to increase by 20–30% in the upcoming 3–5 years. Machine…