H2O-Net: Self-Supervised Flood Segmentation via Adversarial Domain Adaptation and Label Refinement

  title={H2O-Net: Self-Supervised Flood Segmentation via Adversarial Domain Adaptation and Label Refinement},
  author={Peri Akiva and Matthew Purri and Kristin J. Dana and Beth Tellman and Tyler Anderson},
  journal={2021 IEEE Winter Conference on Applications of Computer Vision (WACV)},
Accurate flood detection in near real time via high resolution, high latency satellite imagery is essential to prevent loss of lives by providing quick and actionable information. Instruments and sensors useful for flood detection are only available in low resolution, low latency satellites with region re-visit periods of up to 16 days, making flood alerting systems that use such satellites unreliable. This work presents H2O-Network, a self-supervised deep learning method to segment floods from… 

Figures and Tables from this paper

Self-Supervised Learning for Building Damage Assessment from Large-scale xBD Satellite Imagery Benchmark Datasets

A self-supervised comparative learning approach to address the task without the requirement of labeled data is proposed and a novel asymmetric twin network architecture is constructed and tested its performance on the xBD dataset.

Attentive Dual Stream Siamese U-Net for Flood Detection on Multi-Temporal Sentinel-1 Data

  • Ritu YadavA. NascettiY. Ban
  • Computer Science, Environmental Science
    IGARSS 2022 - 2022 IEEE International Geoscience and Remote Sensing Symposium
  • 2022
This work proposes a segmentation network using bi-temporal SAR data with an encoder-decoder architecture with two Siamese encoders for pre and post-flood images that achieves more accurate fiood detection than uni-Temporal methods.

Towards Single Stage Weakly Supervised Semantic Segmentation

This work utilizes point annotations to generate reliable, on-the-fly pseudo-masks through refined and filtered features, and requires point annotations that are only slightly more expensive than image-level annotations, to demonstrate SOTA performance on benchmark datasets, as well as significantly outperform other SOTA WSSS methods on recent real-world datasets.

Self-Supervised Material and Texture Representation Learning for Remote Sensing Tasks

The effectiveness of the self-supervised pre-trained features in both raw and fine-tuned forms is shown, obtaining state-of-the-art (SOTA) performance in change detection, land cover classification, and semantic segmentation tasks.

Self-Supervised Learning for Invariant Representations From Multi-Spectral and SAR Images

This work proposes RS-BYOL, which builds on BYOL in the remote sensing (RS) domain where data are nontrivially different from natural RGB images, and explores the usefulness of single channel feature learning from random 10 MS bands of 10–20 m resolution and VV-VH of SAR bands compared to the common notion of using three or more bands.

Self-supervised Learning in Remote Sensing: A Review

A preliminary benchmark of modern SSL algorithms on popular remote sensing datasets is provided, verifying the potential of SSL in remote sensing and providing an extended study on data augmentations to pave the way for fruitful interaction of both domains.

Self-Supervised Contrastive Pre-Training For Time Series via Time-Frequency Consistency

A decomposable pre-training model, where the self-supervised signal is provided by the distance between time and frequency components, each individually trained by contrastive estimation is proposed, motivated by TF-C.

Leveraging convolutional neural networks for semantic segmentation of global floods with PlanetScope imagery

Constellations of small satellites are able to achieve the coverage, revisit rate, and spatial resolution needed to generate high resolution global maps of flood waters. The PlanetScope constellation



ADVENT: Adversarial Entropy Minimization for Domain Adaptation in Semantic Segmentation

This work proposes two novel, complementary methods using (i) entropy loss and (ii) adversarial loss respectively for unsupervised domain adaptation in semantic segmentation with losses based on the entropy of the pixel-wise predictions.

S2A: Wasserstein GAN with Spatio-Spectral Laplacian Attention for Multi-Spectral Band Synthesis

This study intends to address synthesis of high resolution multi-spectral satellite imagery using adversarial learning through spatio-spectrals Laplacian attention, and introduces a new cost function for the discriminator based on spatial attention and domain adaptation loss.

DCAN: Dual Channel-wise Alignment Networks for Unsupervised Scene Adaptation

Dual Channel-wise Alignment Networks (DCAN) are presented, a simple yet effective approach to reduce domain shift at both pixel-level and feature-level in deep neural networks for semantic segmentation.

Domain Adaptation for Semantic Segmentation With Maximum Squares Loss

The maximum squares loss prevents the training process being dominated by easy-to-transfer samples in the target domain, and introduces the image-wise weighting ratio to alleviate the class imbalance in the unlabeled target domain.

Unsupervised Intra-Domain Adaptation for Semantic Segmentation Through Self-Supervision

This work proposes a two-step self-supervised domain adaptation approach to minimize the inter-domain and intra-domain gap together and proposes to employ a self- supervised adaptation technique from the easy to the hard subdomain.

Self-Attention Generative Adversarial Networks

The proposed SAGAN achieves the state-of-the-art results, boosting the best published Inception score from 36.8 to 52.52 and reducing Frechet Inception distance from 27.62 to 18.65 on the challenging ImageNet dataset.

Ensemble Knowledge Transfer for Semantic Segmentation

This paper introduces AeroScapes, a new dataset of 3269 images of aerial scenes annotated with dense semantic segmentations, and proposes a simple but effective approach for transferring knowledge from such diverse domains as an ensemble that can be aggregated to significantly improve performance.

StandardGAN: Multi-source Domain Adaptation for Semantic Segmentation of Very High Resolution Satellite Images by Data Standardization

This work standardizes each source and target domains so that all the data have similar data distributions, and uses the standardized data generated by this method to train a classifier and segment the standardized target domain.

CyCADA: Cycle-Consistent Adversarial Domain Adaptation

A novel discriminatively-trained Cycle-Consistent Adversarial Domain Adaptation model that adapts representations at both the pixel-level and feature-level, enforces cycle-consistency while leveraging a task loss, and does not require aligned pairs is proposed.

Sen1Floods11: a georeferenced dataset to train and test deep learning flood algorithms for Sentinel-1

The results suggest deep learning models for flood detection of radar data can outperform threshold based remote sensing algorithms, and perform better with training labels that include flood water specifically, not just permanent surface water.