Best approach for unlabeled data?

I just finished the third lecture and am trying a Kaggle competition as suggested. This challenge however provides a large database of unlabeled data. What is considered the best course of action for this? Does fast.ai library have a way of generating labels from features? It’s a relatively large dataset so I don’t think going through it manually is realistic.

Can you please let me which kaggle problem you are talking about.

This one: https://www.kaggle.com/c/landmark-retrieval-challenge