Getting Started Vision Model Training

In an effort to increase accessibility and exposure to machine learning, astica offers a web interface for training models online. The web interface uses the public REST API which has code published on Github. This page will offer a reference for new users as well as providing a list of recommendations and optimizations for training a vision model.

---

Explore: Vision Model Terminology

Glossary

When training a custom model with astica API or Web UI, there are three main components:

**Classes**: Classes represent the various categories or labels that you want your AI model to recognize. For instance, in an image recognition task, classes could be "dog," "cat," and "other". Each class has a collection of samples that belong to that particular category.

**Samples**: Samples are the individual data points within each class. In the context of image classification, these would be the actual images that belong to a specific class or category. The quality and quantity of the samples directly influence the model's ability to learn and create a generalized understanding of the problem.

**Metadata**: Metadata provides additional information about the dataset, such as the dataset's title, purpose, date of creation, and associated keywords or tags. Metadata helps in organizing and managing your datasets effectively. For custom models trained with astica, these are properties which can be used to discretely enhance the model.

For Developers & App Integration

Custom Model Training API

The custom model training API offers a simple method for training custom artificial intelligence models including vision AI and natural language.

The API offers developers with a convenient and effective way to create, manage, and retrieve datasets required for the training process and to then the train model and begin running inference.

View API Documentation