EddyNet: A Deep Neural Network For Pixel-Wise Classification of Oceanic Eddies

  • 2017-11-10 18:30:05
  • Redouane Lguensat, Miao Sun, Ronan Fablet, Evan Mason, Pierre Tandeo, Ge Chen
  • 5

Abstract

This work presents EddyNet, a deep learning based architecture for automatededdy detection and classification from Sea Surface Height (SSH) maps providedby the Copernicus Marine and Environment Monitoring Service (CMEMS). EddyNet isa U-Net like network that consists of a convolutional encoder-decoder followedby a pixel-wise classification layer. The output is a map with the same size ofthe input where pixels have the following labels \{'0': Non eddy, '1':anticyclonic eddy, '2': cyclonic eddy\}. We investigate the use of SELUactivation function instead of the classical ReLU+BN and we use an overlapbased loss function instead of the cross entropy loss. Keras Python code, thetraining datasets and EddyNet weights files are open-source and freelyavailable on https://github.com/redouanelg/EddyNet.

 

Quick Read (beta)

loading the full paper ...