A “training project” is a smaller project carried out before a “production project”. It’s usually smaller than the production project(s) (or batches), both in size and in duration, and its goal is to reduce uncertainty before them. This type of project allows the annotators to be trained on the tasks, but also to properly configure the labeling tool, apply the correct labeling process and clearly define edge cases before starting production.
We, as People for AI, we refer to a training project when the project is already paid.
The labeled data during the training projects are often of lower quality. The data scientist is free to use them or not to train their model.
A training project and a POC project are synonymous, thus, going through a training project has many advantages like a POC project.
