This dataset consists of several underwater videosequences collected by NCHC in Taiwan and used in the Fish4Knowledge project
The training set is built up of 20 manually annotated videos, a list of 15 fish species and for each species, a set of about 20,000 sample images to support learning
of fish appearance models. The test set contains, instead, 73 underwater videos fully labelled.
Annotations are provided as bounding boxes and fish species name.