I wonder if you still can get meaningful data with some machine learning... But it looks alright.
Guess if I took huge amount of pictures of the same object from different angles maybe some statistical methods could recover transformation and data loss (identical for each photo).