Practice — Personalizing Image Generation: DreamBooth, LoRA and Textual Inversion (6 questions)

A teammate is implementing DreamBooth for the headshot product and proposes two candidate identifiers for the fine-tuning prompt "a photo of [V] person":

• Option A: use the common word "human" as [V].
• Option B: use a randomly generated string, "zqxpj42", as [V].

1. Explain specifically why each option is likely to fail, in mechanistic terms (not just "it's a bad idea").
2. What property should the actual identifier have, and what's the canonical example from the DreamBooth paper?
3. Why does the identifier get paired with a class noun ("person") rather than used alone?

A team trains DreamBooth on 15 photos of a user, using the prompt "a photo of [V] person", with no prior preservation loss (reconstruction loss only). After training, two problems show up:

• Generated images using [V] look almost identical to the exact poses and backgrounds in the 15 training photos, even when the prompt asks for a different setting.
• Generating "a photo of a person" with no identifier now produces images that look suspiciously like the fine-tuned subject too.

1. Name each failure mode and explain the mechanism behind it.
2. Explain what class-specific prior preservation loss adds to training, and where its "generic class" training images come from.
3. Write the combined training objective at the level of detail this subject uses, and explain what the weighting term controls.