A Highly Efficient Marine Mammals Classifier Based on a Cross-Covariance Attended Compact Feed-Forward Sequential Memory Network (Student Abstract)

Military active sonar and marine transportation are detrimental to the livelihood of mammals ecosystem. Early detection classification using machine learning can help humans mitigate harm mammals. This paper proposes a cross-covariance attended compact Feed-Forward Sequential Memory Network (CC-FSMN). The proposed framework shows improved efficiency over multiple convolutional neural network (CNN) backbones. It also maintains relatively decent performance.