The dataset "Manta2018" contains 990 class-labeled ventral Manta alfredi digital images with ROI annotations. Images are covering 99 individuals with exactly 10 images per individual. The dataset was used in the CIARP2018 paper:
T Celik, B Hughes, T Burghardt. "Towards End-to-End DNN-based Identification of Individual Manta Rays from Sparse Imagery" published at the 23rd Iberoamerican Congress on Pattern Recognition (CIARP), November 2018.

The dataset has been provided by:
The Manta Trust
Catemwood House
Corscombe Dorchester
Dorset DT2 0NT
United Kingdom. 

The Manta Trust holds copyrights of all data. 
The University of Bristol holds a copy of the dataset for project documentation purposes only and cannot release the dataset without the copyright holder's explicit permission.

Please contact The Manta Trust directly to obtain the dataset.