A cat-versus-dog classifier is trained on 1,000 photographs for 30 epochs. Every time a photograph is loaded, which happens once per epoch, the augmentation pipeline builds a fresh random version of it by making three independent choices:
| choice | options |
|---|---|
| horizontal flip | flipped or not, each with probability |
| crop | one of crop positions, all equally likely |
| brightness | one of levels, all equally likely |
Every combination of choices produces a different image, and the choices made in one epoch are independent of those made in every other epoch.
Pick any one photograph. Across the 30 epochs, what is the expected number of distinct versions of it that the model sees?
Round your answer to 2 decimal places.