Home

In the data science world, data is often cheap but labels are expensive. In some cases, the data itself is difficult to obtain. This is a perfect scenario for the Flask, which creates artificial data quickly without the need to gather it.

On this page you will find:

How to Augment Data with the Flask

  1. After creating your project and importing your dataset, split your data into training and validation sets. *Learn more about splitting datasets here.*

<aside> <img src="https://s3-us-west-2.amazonaws.com/secure.notion-static.com/52e85db3-0847-4231-a3ed-8afc6f2124ba/0_JAXON_Logo_Mark.jpg" alt="https://s3-us-west-2.amazonaws.com/secure.notion-static.com/52e85db3-0847-4231-a3ed-8afc6f2124ba/0_JAXON_Logo_Mark.jpg" width="40px" /> We recommend running the model at least once without augmented data in order to establish a baseline for comparison.

</aside>

  1. In the Datasets tab, select the dataset you want to augment.

click train set.png

<aside> <img src="https://s3-us-west-2.amazonaws.com/secure.notion-static.com/52e85db3-0847-4231-a3ed-8afc6f2124ba/0_JAXON_Logo_Mark.jpg" alt="https://s3-us-west-2.amazonaws.com/secure.notion-static.com/52e85db3-0847-4231-a3ed-8afc6f2124ba/0_JAXON_Logo_Mark.jpg" width="40px" /> Only training data should be augmented. Polluting validation data with augmented examples may cause the model to appear more accurate than it really is (overfitting). We recommend testing your model exclusively on real data.

</aside>

  1. Click the flask icon to open the data augmentation form.

click flask.png

<aside> <img src="https://s3-us-west-2.amazonaws.com/secure.notion-static.com/52e85db3-0847-4231-a3ed-8afc6f2124ba/0_JAXON_Logo_Mark.jpg" alt="https://s3-us-west-2.amazonaws.com/secure.notion-static.com/52e85db3-0847-4231-a3ed-8afc6f2124ba/0_JAXON_Logo_Mark.jpg" width="40px" /> We recommend giving the new dataset a descriptive name to avoid confusion in the future.

</aside>

  1. Adjust the augmentation settings to your preference. More details about the settings can be found below.

flask-screenshot-3.png

  1. Select Submit . Jaxon will preserve the original dataset and create a new one that includes the augmented examples.

flask-screenshot-arrow.png

Back to the top ↑

  1. The new augmented dataset will appear in the Datasets tab under the name you selected for it.

Data Augmentation: Form Breakdown

Back to the top ↑

© Copyright Jaxon, Inc. 2023 All rights reserved.