The aim of this competition is to build a machine learning model to identify individual sea turtles. For each image presented, the model should output the turtle’s unique ID or, if the image corresponds to a new turtle (not present in the database), identify it as a new individual. A dataset of labelled photos of turtle faces and a tutorial including a simple baseline model have been provided.
The model used in this submission is ResNet152 with ImageNet policies and heavy image augmentaiton.