You’ll receive a dataset of uncaptioned images.
There is no ground truth provided

Hint: The better your captions, the more meaningful your CLIP embeddings. Focus on generating captions that are descriptive, relevant, and varied.