Brain Builder » Recognition » Recognition Documentation » Uploading Data for Recognition Training

In This Section:

On This Page:

Before you begin...

Read the following tips to ensure that your data is as effective as possible.

Recognition Brains look at the whole image and determine if an object on which they have been trained is present.

When selecting data for a Recognition Brain, it's important that:

  • The objects in the image are prominent and clearly visible.
  • Each image contains a single target object class. (The term "class" refers to the label applied to the image.)
  • The classes are distinct from one another.

Quality over quantity.

Brain Builder begins to learn after you tag a single image, and it can usually begin making predictions after just five or ten. Unlike traditional vision AI systems, Brain Builder works with less data.

It's far better to have a smaller dataset with consistently good images than a larger dataset with troubling images. If you're not sure how much data you'll need, start with a few dozen images for each class. Because Brain Builder learns incrementally, you can always add more data later on.

Uploading Data

After you have created your Recognition dataset in Creating a Project, it’s time to upload images.

  1. From the Dataset page, click the Upload Images button in the top-right corner.

  2. Add a tag for the batch if you’re uploading images that are all of the same class. The tag would be the class name.

    If your images are already organized by class, uploading each class as a separate batch and adding a tag will save you time because you won’t have to go through and tag each image individually.

    If they are not already organized by class, you can upload images of different classes at the same time and not add a tag.

  3. Click either Upload from Computer or Upload image from URL to select the photos you want to upload.

    Warning URLs must be publicly reachable so that Brain Builder may retrieve the data files. Files behind firewalls or on corporate intranets may not work.

    Neurala Brain Builder supports the following common image file formats:

    • GIF
    • JPG/JPEG
    • PNG
    • SVG
    • TIFF

    You may also upload a ZIP file containing multiple images.

    Information Images may not be more than 16,000 pixels tall or wide.
  4. Click Start Upload.

    The progress bar shows you the status of your data uploads while the text below displays a count of the number of items that upload successfully or fail to process.

    Information Data may fail to upload for a number of reasons. If your images do not upload successfully, please review the list of supported data types and data limitations.

Next: Training your Brain >>