Improving Right Whale recognition by fine-tuning alignment and using wide localization network

Abstract

Right Whales can be recognized by the callosities pattern on their heads. They are an endangered species with an estimated 450 whales remaining. Marine biologists regularly perform manual recognition of the whales while monitoring the population but the process is slow and time consuming. Deep learning methods achieved state-of-the-art results on several visual recognition tasks. However, training deep learning models on this task is very difficult because the number of training images is low. We propose a wide localization network which can be used to localize the region of interest in image. Once the region of interest is localized, a deep learning model can be used to classify the whales. The solution we describe in this paper achieves an accuracy score of 78.7% and ranks as one of the best 3 solutions on this dataset.

DOI: 10.1109/CCECE.2017.7946736

7 Figures and Tables

Cite this paper

@article{Kabani2017ImprovingRW, title={Improving Right Whale recognition by fine-tuning alignment and using wide localization network}, author={AbdulWahab Kabani and Mahmoud R. El-Sakka}, journal={2017 IEEE 30th Canadian Conference on Electrical and Computer Engineering (CCECE)}, year={2017}, pages={1-6} }