- the *_labels.json files contain annotations for all corresponding image files
- some files contain more than one fish - only the most dominant fish was annotated
- the first point is always for the head, the second point is for the tail
- the annotations were made with the idea of cropping the fish prior to learning its type -
- suggested margins for the crops are 10% for the long axis and 50% to each of the sides (creating a square)
- if the fish was partly hidden the last seen spot was annotated