Brain Builder Knowledge Base
What's New in Brain Builder
Direct link to topic in this publication:
Brain Builder: Recognition End-to-End
The latest release of the Brain Builder platform, along with our proprietary Lifelong-DNN™ technology, allows AI beginners and experts alike to create custom vision AI solutions that solve business problems faster, less expensively, and on a secure platform. With a lot less data and without requiring special hardware, anyone can build a robust, deployable AI solution using nothing but a personal computer.
We are also releasing our Knowledge Base, a Getting Started Guide for Recognition Use Cases, and Getting Started Walkthroughs.
With this release of Brain Builder, you can train a custom Recognition model and deploy it on the cloud. Get started by creating a Recognition dataset, upload some images, annotate them, and deploy.
The model learns and predicts classifications as you tag data. When you create a Recognition dataset, every image that you tag will be immediately used to train a model. And, if we have enough data, we use this model to predict the classification of the next image to make it easier for you and demonstrate how well the model is working at that point.
The Brain Score feature allows users of any skill level to quickly understand how well their current brain is working. We use factors such as Number of Images per Class, Relative Class Accuracy, and Overall Accuracy to determine the Brain Score. Also, when clicked, the Brain Score can give you additional insights such as what classes need more images and what are the images that are being predicted incorrectly by the model.
Because we understand that you want to deploy an AI solution ASAP, but would also like to continue improving the model, we are releasing the ability to version your brains into:
Current—the brain that you are actively training
Staging—a version of the Brain that you’d like to test in the field but may not be ready for production
Production—the brain that most, if not all, your applications should be running in the field
Once you promote a Brain to Staging or Production, you can run inferences in the cloud by using our APIs. Also, we have released a way to manage your API keys in case you’d like to have multiple keys for multiple products.