The AQUA20 dataset is a comprehensive benchmark dataset designed for underwater species classification under challenging real-world conditions. It comprises 8,171 underwater images across 20 distinct marine species, specifically curated to reflect environmental complexities such as turbidity, low illumination, and occlusion, which commonly degrade the performance of standard vision systems. This dataset provides a valuable resource for advancing robust visual… See the full description on the dataset page:
https://huggingface.co/datasets/Mrscarce/AQUA20.
