Quantifying Intertidal Zone Species Using Semantic Segmentation

As anthropogenic impacts on marine ecosystems accelerates (e.g. warming, acidification, eutrophication, etc), it is essential to build robust datasets that establish biological baseline data and capture long-term trends in shifting species abundance and diversity. This data has traditionally been collected through continual revisits by skilled ecologists and taxonomists to long-term ecological monitoring sites. One novel technique developed by an intertidal ecology research group at California State University Channel Islands (CSUCI) builds 1m-wide photo-transects for the length of the tidal zone (20m from splash to low zone) at two sites on Santa Rosa Island. These photos are stitched together using software and offer high-resolution swaths of information at the island, taken twice a year. A machine learning technique, semantic segmentation, has been employed to automate the analysis of these large images, focusing first on a dominant algal species of rockweed Silvetia compressa. This automation will greatly reduce the time needed and human error involved in scoring and quantifying these transects. The study involves developing a convolutional neural network using transfer learning on a publicly available network.

[1]  Nitish Srivastava,et al.  Dropout: a simple way to prevent neural networks from overfitting , 2014, J. Mach. Learn. Res..

[2]  Trevor Darrell,et al.  Transfer learning for image classification with sparse prototype representations , 2008, 2008 IEEE Conference on Computer Vision and Pattern Recognition.

[3]  Vladlen Koltun,et al.  Multi-Scale Context Aggregation by Dilated Convolutions , 2015, ICLR.

[4]  Yoshua Bengio,et al.  How transferable are features in deep neural networks? , 2014, NIPS.

[5]  Heather M. Leslie,et al.  Applying ecological criteria to marine reserve design: A case study from the california channel islands , 2003 .

[6]  Iasonas Kokkinos,et al.  DeepLab: Semantic Image Segmentation with Deep Convolutional Nets, Atrous Convolution, and Fully Connected CRFs , 2016, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[7]  George Papandreou,et al.  Rethinking Atrous Convolution for Semantic Image Segmentation , 2017, ArXiv.

[8]  Rajat Raina,et al.  Self-taught learning: transfer learning from unlabeled data , 2007, ICML '07.

[9]  David Rousseau,et al.  Transfer Learning from Synthetic Data Applied to Soil-Root Segmentation in X-Ray Tomography Images , 2018, J. Imaging.

[10]  Andrew W. Fitzgibbon,et al.  Real-time human pose recognition in parts from single depth images , 2011, CVPR 2011.

[11]  Roberto Cipolla,et al.  SegNet: A Deep Convolutional Encoder-Decoder Architecture for Image Segmentation , 2015, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[12]  Thomas Brox,et al.  U-Net: Convolutional Networks for Biomedical Image Segmentation , 2015, MICCAI.

[13]  Arati Dandavate,et al.  Semantic Texton Forests for Image Categorization and Segmentation , 2018, IJARCCE.