4.2 What things should I pay attention to about the Ground-Truth data?

For participants of the Federated Validation project, the way to provide Ground-Truth Dataset(s) must be of the formats, including the other requests, like labeling names, column naming, file naming, folder naming, etc., which meet the requirements of validating AI model.

Usually, all datasets are compressed as a single file to upload to the FV Edge Machine, to make sure the folder arrangement won't get messed up during the dataset transmission. The compressed file will be uncompressed automatically, and with the protocol that FV System designs for AI Model, the datasets will then be read and processed by AI Model docker image afterwards.

If the uploaded datasets don't meet the requirements of validating AI Model, then relative error message(s) will show up on the FV Edge Machine. Please follow the prompt to adjust the datasets.