NeMO-Net The Neural Multi-Modal Observation & Training Network for Global Coral Reef Assessment
We present NeMO-Net, the Srst open-source deep convolutional neural network (CNN) and interactive learning and training software aimed at assessing the present and past dynamics of coral reef ecosystems through habitat mapping into 10 biological and physical classes. Shallow marine systems, particul...
Main Authors: | , , , , , , |
---|---|
Format: | Other/Unknown Material |
Language: | unknown |
Published: |
2020
|
Subjects: | |
Online Access: | http://hdl.handle.net/2060/20200001640 |
id |
ftnasantrs:oai:casi.ntrs.nasa.gov:20200001640 |
---|---|
record_format |
openpolar |
spelling |
ftnasantrs:oai:casi.ntrs.nasa.gov:20200001640 2023-05-15T17:51:56+02:00 NeMO-Net The Neural Multi-Modal Observation & Training Network for Global Coral Reef Assessment Van Den Bergh, Jarrett S. Torres-Perez, Juan Chirayath, Ved Segal-Rozenhaimer, Michal Li, Alan Das, Kamalika Purkis, Sam Unclassified, Unlimited, Publicly available February 16, 2020 application/pdf http://hdl.handle.net/2060/20200001640 unknown Document ID: 20200001640 http://hdl.handle.net/2060/20200001640 Copyright, Public use permitted CASI Earth Resources and Remote Sensing ARC-E-DAA-TN73997 Ocean Sciences Meeting 2020; Feb 16, 2020 - Feb 21, 2020; San Diego, CA; United States 2020 ftnasantrs 2020-03-21T23:47:53Z We present NeMO-Net, the Srst open-source deep convolutional neural network (CNN) and interactive learning and training software aimed at assessing the present and past dynamics of coral reef ecosystems through habitat mapping into 10 biological and physical classes. Shallow marine systems, particularly coral reefs, are under significant pressures due to climate change, ocean acidification, and other anthropogenic pressures, leading to rapid, often devastating changes, in these fragile and diverse ecosystems. Historically, remote sensing of shallow marine habitats has been limited to meter-scale imagery due to the optical effects of ocean wave distortion, refraction, and optical attenuation. NeMO-Net combines 3D cm-scale distortion-free imagery captured using NASA FluidCam and Fluid lensing remote sensing technology with low resolution airborne and spaceborne datasets of varying spatial resolutions, spectral spaces, calibrations, and temporal cadence in a supercomputer-based machine learning framework. NeMO-Net augments and improves the benthic habitat classification accuracy of low-resolution datasets across large geographic ad temporal scales using high-resolution training data from FluidCam.NeMO-Net uses fully convolutional networks based upon ResNet and ReSneNet to perform semantic segmentation of remote sensing imagery of shallow marine systems captured by drones, aircraft, and satellites, including WorldView and Sentinel. Deep Laplacian Pyramid Super-Resolution Networks (LapSRN) alongside Domain Adversarial Neural Networks (DANNs) are used to reconstruct high resolution information from low resolution imagery, and to recognize domain-invariant features across datasets from multiple platforms to achieve high classification accuracies, overcoming inter-sensor spatial, spectral and temporal variations.Finally, we share our online active learning and citizen science platform, which allows users to provide interactive training data for NeMO-Net in 2D and 3D, integrated within a deep learning framework. We present results from the PaciSc Islands including Fiji, Guam and Peros Banhos 1 1 2 1 3 1 where 24-class classification accuracy exceeds 91%. Other/Unknown Material Ocean acidification NASA Technical Reports Server (NTRS) Pyramid ENVELOPE(157.300,157.300,-81.333,-81.333) |
institution |
Open Polar |
collection |
NASA Technical Reports Server (NTRS) |
op_collection_id |
ftnasantrs |
language |
unknown |
topic |
Earth Resources and Remote Sensing |
spellingShingle |
Earth Resources and Remote Sensing Van Den Bergh, Jarrett S. Torres-Perez, Juan Chirayath, Ved Segal-Rozenhaimer, Michal Li, Alan Das, Kamalika Purkis, Sam NeMO-Net The Neural Multi-Modal Observation & Training Network for Global Coral Reef Assessment |
topic_facet |
Earth Resources and Remote Sensing |
description |
We present NeMO-Net, the Srst open-source deep convolutional neural network (CNN) and interactive learning and training software aimed at assessing the present and past dynamics of coral reef ecosystems through habitat mapping into 10 biological and physical classes. Shallow marine systems, particularly coral reefs, are under significant pressures due to climate change, ocean acidification, and other anthropogenic pressures, leading to rapid, often devastating changes, in these fragile and diverse ecosystems. Historically, remote sensing of shallow marine habitats has been limited to meter-scale imagery due to the optical effects of ocean wave distortion, refraction, and optical attenuation. NeMO-Net combines 3D cm-scale distortion-free imagery captured using NASA FluidCam and Fluid lensing remote sensing technology with low resolution airborne and spaceborne datasets of varying spatial resolutions, spectral spaces, calibrations, and temporal cadence in a supercomputer-based machine learning framework. NeMO-Net augments and improves the benthic habitat classification accuracy of low-resolution datasets across large geographic ad temporal scales using high-resolution training data from FluidCam.NeMO-Net uses fully convolutional networks based upon ResNet and ReSneNet to perform semantic segmentation of remote sensing imagery of shallow marine systems captured by drones, aircraft, and satellites, including WorldView and Sentinel. Deep Laplacian Pyramid Super-Resolution Networks (LapSRN) alongside Domain Adversarial Neural Networks (DANNs) are used to reconstruct high resolution information from low resolution imagery, and to recognize domain-invariant features across datasets from multiple platforms to achieve high classification accuracies, overcoming inter-sensor spatial, spectral and temporal variations.Finally, we share our online active learning and citizen science platform, which allows users to provide interactive training data for NeMO-Net in 2D and 3D, integrated within a deep learning framework. We present results from the PaciSc Islands including Fiji, Guam and Peros Banhos 1 1 2 1 3 1 where 24-class classification accuracy exceeds 91%. |
format |
Other/Unknown Material |
author |
Van Den Bergh, Jarrett S. Torres-Perez, Juan Chirayath, Ved Segal-Rozenhaimer, Michal Li, Alan Das, Kamalika Purkis, Sam |
author_facet |
Van Den Bergh, Jarrett S. Torres-Perez, Juan Chirayath, Ved Segal-Rozenhaimer, Michal Li, Alan Das, Kamalika Purkis, Sam |
author_sort |
Van Den Bergh, Jarrett S. |
title |
NeMO-Net The Neural Multi-Modal Observation & Training Network for Global Coral Reef Assessment |
title_short |
NeMO-Net The Neural Multi-Modal Observation & Training Network for Global Coral Reef Assessment |
title_full |
NeMO-Net The Neural Multi-Modal Observation & Training Network for Global Coral Reef Assessment |
title_fullStr |
NeMO-Net The Neural Multi-Modal Observation & Training Network for Global Coral Reef Assessment |
title_full_unstemmed |
NeMO-Net The Neural Multi-Modal Observation & Training Network for Global Coral Reef Assessment |
title_sort |
nemo-net the neural multi-modal observation & training network for global coral reef assessment |
publishDate |
2020 |
url |
http://hdl.handle.net/2060/20200001640 |
op_coverage |
Unclassified, Unlimited, Publicly available |
long_lat |
ENVELOPE(157.300,157.300,-81.333,-81.333) |
geographic |
Pyramid |
geographic_facet |
Pyramid |
genre |
Ocean acidification |
genre_facet |
Ocean acidification |
op_source |
CASI |
op_relation |
Document ID: 20200001640 http://hdl.handle.net/2060/20200001640 |
op_rights |
Copyright, Public use permitted |
_version_ |
1766159226978172928 |