Create a task by specifying input modality and output concepts. For every concept submitted, you will be required to provide a few relevant examples. Then pre-training will begin and you can start integrating XPL within your software.
Once the model is trained, you will be able to integrate it into your project with just a few lines of code. We recommend flagging the feature and immediately deploying your software - so that your task gets exposed to real-world data. The task will analyze the data it is exposed to and collect a small subset of useful samples.
After your code is deployed, you can monitor how your task performs on real-world data. Meanwhile, the XPL Platform analyzes collected data and automatically creates annotation jobs from a subset of the most informative samples. You are notified when new annotation jobs are created. XPL Annotation UI is fine-grained for usability, requiring the absolute minimal amount of effort. Annotating will trigger a new cycle of training and your task’s performance will go up. There is no need to redeploy your software as rollout logic is already included in the client integration component.