CLIP (Contrastive Language-Image Pretraining), Predict the most relevant text snippet given an image
Labels help you categorize and filter issues.