Comparison of deep learning approaches in identifying humpback whales based on fin images. The dataset used for was obtained from Kaggle. It contains thousands of images of humpback whale flukes. Researchers have identified individual whales and given an identification (Id). The challenge is to predict the whale Id of images in the test set.
We used a CNN, Transfer Learning, and a SVM and compare their results to find the most effective model for the identification of individual humpback whales. CNN and Transfer learning predictably perform well, but due to overfitting in the CNN model, Transfer Learning is the better model.