Skip to content
View MARDAScience's full-sized avatar
23 followers · 13 following

Achievements

Achievement: StarstruckAchievement: Arctic Code Vault Contributor

Achievements

Achievement: StarstruckAchievement: Arctic Code Vault Contributor

Block or report MARDAScience

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Please don't include any personal information such as legal names or email addresses. Maximum 100 characters, markdown supported. This note will be visible to only you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
MARDAScience/README.md

Marda Science

We are a tiny consultancy in environmental analytics and research, specializing in coastal, marine, and river environments.

  • We specialize in ML/Geospatial/Algorithm/Software for environmental science, engineering and management.
  • We conduct research and development of novel measurements and insights from data-driven analyses involving artificial intelligence, computer vision, and/or numerical modeling.
  • We develop scalable automated workflows for extracting information from a plethora of environmental, geophysical, oceanographic, geological and ecological data.

We consist of Dr Daniel Buscombe and Dr Maria Campbell. This is the landing page for Dan's page of software repositories and datasets that concern projects in Machine Learning, Deep Learning, Remote Sensing, Computer Vision, Image Processing, Sonar Processing, Numerical Modeling, and Geospatial analysis. It gives a more complete picture than our website. Some teaching resources are hosted through our Gitlab repository. See also Dan's youtube page. Dan's other github handle is dbuscombe-usgs. He often develop codes collaboratively and host them mostly in a range of Github organizations, listed below:

Active:

Inactive:

Contents

Media

November 2022: CSDMS webinar.

Part 1 of the 2-part, ``Intro to the Doodleverse'' webinar for the Community Surface Dynamics Modeling System fall webinar series youtube link concentrates on Doodler

Part 2 of the 2-part, ``Intro to the Doodleverse'' webinar for the Community Surface Dynamics Modeling System fall webinar series youtube link concentrates on Segmentation Gym

November 2022: satellite-image-deep-learning Podcast

Interview with Robin Cole for the satellite-image-deep-learning podcast and newsletter entitled, ``Into the Doodleverse''

August 2022: Bodega Bay talk on applications of Deep Learning for coastal monitoring

Talk entitled "Developing Deep Learning Design Patterns for Large-Scale Coastal Monitoring", for the John & Mary Louise Riley Bodega Marine Laboratory Seminar Series, UC Davis Bodega Marine Lab, Bodega Bay, CA.

December 2021: The International Society for Photogrammetry and Remote Sensing Student Consortium (ISPRS SC)

Interview (5 pages) in SpeCtrum (Vol. 15, No. 2), the official newsletter of the The International Society for Photogrammetry and Remote Sensing Student Consortium (ISPRS SC)

March 2016: USGS "What's the Big Idea?""

My acoustic remote sensing research was featured in the video on the YouTube channel of the U.S. Geological Survey

July 2015: American Geophysical Union Research Spotlight

My acoustic remote sensing research featured in EOS Earth and Space Science News

September 2012]{JGR-Oceans Editor's Highlight

My oceanographic research was featured in an article published in the Journal of Geophysical Research - Oceans. *[Novel observations of currents and drag generated by a tsunami](http://agupubs.onlinelibrary.wiley.com/agu/article/10.1029/2012JC007954/editor-highlight/}

Active Software Projects

RUSH

RUSH: Rapid Remote Sensing Updates of landcover for Storm and Hurricane forecasts

Written by Winston Cheang, USGS WGSC | [email protected], with support from Daniel Buscombe and others. See the software available here

SDSTools

A python toolkit for SDS data post-processing, analysis, and visualization.

In the past year I have overseen a small team developing satellite-image based shoreline mapping tools as we work towards a prospectus outlined in a recent paper.

Doodleverse

Deep-learning based semantic segmentation of geospatial data.

In the past two years I led the development and now maintain a set of TensorFlow-based tools specifically designed for this task - from developing training data to creating deployment ready models. The set of tools is available on Github Doodleverse Org.. I have recently talked about these tools and how they are being applied in production.

That work has spawned the development of several downstream applications for specific tasks. For example, CoastSeg and Seg2Map described below

CoastSeg

An interactive interface to download satellite imagery using CoastSat from Google Earth Engine, extracting shorelines from satellite imagery, and applying segmentation models to satellite imagery A mapping extension for CoastSat using Segmentation Zoo models

In the past year I have overseen a small team developing satellite-image based shoreline mapping tools as we work towards a prospectus outlined in a recent paper.

Ping-mapper

Sidescan sonar processing and analysis

I have been involved in sidescan sonar processing for a decade. I oversee a software project for reading, processing and analysis of Humminbird sidescan data, called [Ping-mapper]https://github.com/CameronBodine/PINGMapper/). It is based on older software I wrote for for reading, processing and analysis of Humminbird sidescan data. Source code available in Python/Cython here is now archived, having been succeeded by PING-Mapper.

RegionalGrainSizeModel

Machine Learning for estimating beach grain size over regional scales

This ongoing project uses boosted regression trees to estimate sand beach grain size over regional scales from a suite of covariates like beach slope, tide, and wave climate.

Inactive Software Projects

Seg2Map

An interactive web map app for geospatial label imagery generated within the Doodleverse

Another generic toolbox for semantic segmentation of geospatial imagery is Seg2Map. Display the imagery on a web map, and apply segmentation models to create labels and maps

SediNet

Deep-learning based estimation of sediment grain size

Several of my model frameworks, including SediNet, combine to amass a citizen science database of beach sand measurements for the U.S. Army Corps of Engineers, called the SandSnap project. SediNet is software for application of deep convolutional neural networks to estimation of quantitative and qualitative properties of sediment in photographic imagery.

Other (older) software for automated analyses of grain size from images of sediment usign wavelets is also available and widely used, having amassed several hundred academic paper citations.

OWG

Deep-learning based estimation of nearshore ocean wave properties

I research methods to measure surf zone waves from satellite imagery by adapting a deep learning model framework I developed to larger scales.

Software for application of deep convolutional neural networks to estimation of ocean wave properties from time-series of imagery are available in the Github Optical Wave Gauging Org.

RetinaDamage

Deep Learning for estimating damage to buildings due to natural disasters

This project trained a RetinaNet model to detect building damage in Maxar satellite imagery.

GobyNet

Deep-learning based detection of benthic fish

I have written software for automated detection of camouflaged benthic fish using models based on RetinaNet and deep residual U-Net. Source code currently available in Python (link soon).

prism

Multibeam processing and analysis

Software for probabilistic seafloor habitat mapping using multibeam backscatter. Code here

PBR-filter

Novel {P}ansharpening by {B}ackground {R}emoval algorithm for sharpening RGB images. Code here

pysesa

Software for spatially explicit analyis of point clouds and spatially distributed data. Code (now archived) is here.

Teaching materials

I have taught several classes on Machine Learning and data science principles, including development and teaching of Machine Learning courses for my coworkers and the wider geoscience community.

ML-Mondays

Teaching materials and software for application of deep convolutional neural networks for analysis of photographic imagery. Supervised and semi-supervised deep learning models and data for image segmentation, and object detection. Source code currently available in Python on the ML-Mondays Github Org.

Image Segmentation with UNets

Published datasets and models

Doodleverse Zenodo releases

ML models

Current Seg2Map models
Generic landcover

  • Buscombe, Daniel. (2023). Doodleverse/Segmentation Zoo/Seg2Map Res-UNet models for FloodNet/10-class segmentation of RGB 768x512 UAV images (v1.0) [Data set]. Zenodo. https://doi.org/10.5281/zenodo.7566810

  • Buscombe, Daniel. (2023). Doodleverse/Segmentation Zoo/Seg2Map Res-UNet models for OpenEarthMap/9-class segmentation of RGB 512x512 high-res. images (v1.0) [Data set]. Zenodo. https://doi.org/10.5281/zenodo.7576894

  • Buscombe, Daniel. (2023). Doodleverse/Segmentation Zoo/Seg2Map Res-UNet models for DeepGlobe/7-class segmentation of RGB 512x512 high-res. images (v1.0) [Data set]. Zenodo. https://doi.org/10.5281/zenodo.7576898

  • Buscombe, Daniel. (2023). Doodleverse/Segmentation Zoo/Seg2Map Res-UNet models for EnviroAtlas/6-class segmentation of RGB 512x512 high-res. images (v1.0) [Data set]. Zenodo. https://doi.org/10.5281/zenodo.7576909

  • Buscombe, Daniel. (2023). Doodleverse/Segmentation Zoo/Seg2Map Res-UNet models for FloodNet/10-class segmentation of RGB 1024x768 UAV images (v1.0) [Data set]. Zenodo. https://doi.org/10.5281/zenodo.7566797

Coastal landcover

  • Buscombe, Daniel. (2023). Doodleverse/Segmentation Zoo/Seg2Map Res-UNet models for CoastTrain/5-class segmentation of RGB 768x768 NAIP images (v1.0) [Data set]. Zenodo. https://doi.org/10.5281/zenodo.7566992

  • Buscombe, Daniel. (2023). Doodleverse/Segmentation Zoo/Seg2Map Res-UNet models for CoastTrain/8-class segmentation of RGB 768x768 NAIP images (v1.0) [Data set]. Zenodo. https://doi.org/10.5281/zenodo.7570583

  • Buscombe, Daniel. (2023). Doodleverse/Segmentation Zoo/Seg2Map Res-UNet models for Chesapeake/7-class segmentation of RGB 512x512 high-res. images (v1.0) [Data set]. Zenodo. https://doi.org/10.5281/zenodo.7576904

  • Buscombe, Daniel. (2023). Doodleverse/Segmentation Zoo/Seg2Map Res-UNet models for CoastTrain water/other segmentation of RGB 768x768 orthomosaic images (v1.0) [Data set]. Zenodo. https://doi.org/10.5281/zenodo.7574784

Current CoastSeg models
Older CoastSeg models

  • Buscombe, D. (2022). Doodleverse/Segmentation Zoo Res-UNet models for 4-class (water, whitewater, sediment and other) segmentation of Sentinel-2 and Landsat-7/8 7-band (RGB+NIR+SWIR+NDWI+MNDWI) images of coasts. [Data set]. Zenodo. link

  • Buscombe, D. (2022). Doodleverse/Segmentation Zoo Res-UNet models for 4-class (water, whitewater, sediment and other) segmentation of Sentinel-2 and Landsat-7/8 1-band MNDWI images of coasts. (v1.0) [Data set]. Zenodo. link

  • Buscombe, D. (2022). Doodleverse/Segmentation Zoo Res-UNet models for 4-class (water, whitewater, sediment and other) segmentation of Sentinel-2 and Landsat-7/8 1-band NDWI images of coasts. (v1.0) [Data set]. Zenodo. link

  • Buscombe, D. (2022). Doodleverse/Segmentation Zoo Res-UNet models for 4-class (water, whitewater, sediment and other) segmentation of Sentinel-2 and Landsat-7/8 5-band (RGB+NIR+SWIR) images of coasts. (v1.0) [Data set]. Zenodo. link

  • Buscombe, D. (2022). Doodleverse/Segmentation Zoo Res-UNet models for 4-class (water, whitewater, sediment and other) segmentation of Sentinel-2 and Landsat-7/8 3-band (RGB) images of coasts. (v1.0) [Data set]. Zenodo. link

  • Buscombe, Daniel. (2023). Doodleverse/Segmentation Zoo Res-UNet models for 2-class (water, other) segmentation of Sentinel-2 and Landsat-7/8 1-band MNDWI images of coasts. (v1.0) [Data set]. Zenodo. link

  • Buscombe, Daniel. (2023). Doodleverse/Segmentation Zoo Res-UNet models for 2-class (water, other) segmentation of Sentinel-2 and Landsat-7/8 1-band NDWI images of coasts. (v1.0) [Data set]. Zenodo. link

  • Buscombe, Daniel. (2022). Doodleverse/Segmentation Zoo Res-UNet models for 2-class (water, other) segmentation of Sentinel-2 and Landsat-7/8 3-band (RGB) images of coasts. (v1.0) [Data set]. Zenodo. link

  • Buscombe, Daniel. (2022). Doodleverse/Segmentation Zoo Res-UNet models for 2-class (water, other) segmentation of Sentinel-2 and Landsat-7/8 5-band (RGB+NIR+SWIR) images of coasts. (v1.0) [Data set]. Zenodo. link

  • Buscombe, D. (2022) Segmentation Zoo UNet models for Landsat-8 satellite imagery, Coast Train v1 Landsat-8 4-class subset. (v1.0.0) [Data set]. Zenodo. link

  • Buscombe, D. (2022) Segmentation Zoo Res-UNet models for Landsat-8 satellite imagery, Coast Train v1 Landsat-8 4-class subset. (v1.0.0) [Data set]. Zenodo. link

SandSnap models
  • Buscombe, D. (2022) Doodleverse/Segmentation Zoo Res-UNet models for identifying coins in photos of sediment. (v1.0.0) [Data set]. Zenodo. link
Watermasking models

  • Buscombe, D. (2022) Doodleverse/Segmentation Zoo UNet models for identifying water in oblique aerial photos of coasts. (v1.0.0) [Data set]. Zenodo. link

  • Buscombe, D. (2022) Doodleverse/Segmentation Zoo Res-UNet models for identifying water in oblique aerial photos of coasts. (v1.0.0) [Data set]. Zenodo. link

Labeled imagery for segmentation models

  • Buscombe, D., Lundine, M.A., Janda, C.N., & Batiste, S. (2025) Labeled satellite imagery for training machine learning semantic segmentation models of coastal shorelines. U.S. Geological Survey Data Release 10.5066/P13EOBZQ link

  • Buscombe, D. (2023) June 2023 Supplement Images and 4-class labels for semantic segmentation of Sentinel-2 and Landsat RGB, NIR, and SWIR satellite images of coasts (water, whitewater, sediment, other) (v1.0) [Data set]. Zenodo. link

  • Buscombe, D. (2022). Images and 2-class labels for semantic segmentation of Sentinel-2 and Landsat RGB, NIR, and SWIR satellite images of coasts (water, other) (v1.0) [Data set]. Zenodo. link

  • Buscombe, D. (2022). Images and 2-class labels for semantic segmentation of Sentinel-2 and Landsat RGB satellite images of coasts (water, other) (v1.0) [Data set]. Zenodo. link

  • Buscombe, D. (2022). Images and 4-class labels for semantic segmentation of Sentinel-2 and Landsat RGB, NIR, and SWIR satellite images of coasts (water, whitewater, sediment, other) (v1.0) [Data set]. Zenodo. link

  • Buscombe, D, Goldstein, E, Bernier, J., Bosse, S., Colacicco, R., Corak, N., Fitzpatrick, S., del Jesús González Guillén, A., Ku, V., Paprocki, J., Platt, L., Steele, B., Wright, K., & Yasin, B. (2022). Images and 4-class labels for semantic segmentation of Sentinel-2 and Landsat RGB satellite images of coasts (water, whitewater, sediment, other) (v1.0) [Data set]. Zenodo. link

  • Buscombe, D. (2022). Cape Hatteras Landsat8 RGB Images and Labels for Image Segmentation using the program, Segmentation Zoo (v3.0) [Data set]. Zenodo. link

  • Buscombe, D., et al. (2022) Labeled Images of Sand and Coins v2 (v1.0.0) [Data set]. Zenodo. link

  • Goldstein, E. B., et al. (2022) Segmentation Labels for Emergency Response Imagery from Hurricane Barry, Delta, Dorian, Florence, Isaias, Laura, Michael, Sally, Zeta, and Tropical Storm Gordon (Version v1) [Data set]. Zenodo. link

  • Lima et al. (2020) Semantic Segmentation of Time Series Imagery Using Deep Convolutional Neural Networks: A Case Study of Sandbars in Grand Canyon (Version 2) [Data set]. Zenodo. https://doi.org/10.5281/zenodo.4302747 link

Other Zenodo dataset releases

  • Buscombe, D., & Gómez de la Peña, E. (2025) "Tiny deep learning" model used for sand beach shoreline forecasting (v1.0). Zenodo. https://doi.org/10.5281/zenodo.15003139 link

  • Buscombe, D. (2024) Doodleverse Segformer models for 2-class (scalebar and other) segmentation of visible-band (RGB) images of coastal sediments. (v1.0). Zenodo. https://doi.org/10.5281/zenodo.14502652 \href{https://doi.org/10.5281/zenodo.14502652}{link}}

  • Buscombe, D., Campbell, M. (2024) Median Grain Diameter of Beaches in the Pacific North West, February 2021 to December 2023 (v1.0) [Data set]. Zenodo. https://doi.org/10.5281/zenodo.10467331 link

  • Buscombe, D., Fitzpatrick, S. (2024) CoastSeg: Beach transects and beachface slope database v1.0 (v1.0) [Data set]. Zenodo. https://doi.org/10.5281/zenodo.8187949 link

  • Buscombe, D. (2023) CoastSeg: 30-m atlas of the coastal shoreline attributes of California, in geoJSON format. (v1.0) [Data set]. Zenodo. https://doi.org/10.5281/zenodo.8180524 link

  • Buscombe, D. (2023) Labeled high-resolution orthoimagery time-series of an alluvial river corridor; Elwha River, Washington, USA. (1.0) [Data set]. Zenodo. https://doi.org/10.5281/zenodo.10155783 link

  • Buscombe, D. (2022) Shoreline data at 30-m spatial resolution for 2001 coastal provinces or regions of the world, in geoJSON format. (v1.0.0) [Data set]. Zenodo. https://doi.org/10.5281/zenodo.6917963 link

  • Buscombe, D. et al. (2022) Preliminary Coastal Grain Size Portal (C-GRASP) dataset. Version 1, January 2022, Zenodo https://doi.org/10.5281/zenodo.6099266 link

  • McFall, M., et al. (2022) The SandSnap Project: 2020 -- 2021 sieved grain-size data and associated sediment imagery (0.0.1) [Data set]. Zenodo. link

  • Goldstein, E. B., et al. (2022) Labels for Emergency Response Imagery from Hurricane Barry, Delta, Dorian, Florence, Ida, Isaias, Laura, Michael, Sally, Zeta, and Tropical Storm Gordon (2.0) [Data set]. Zenodo. link

  • Buscombe, D., et al. (2022) Monthly CDIP:MOP-alongshore modeled wave statistics for California, January 2000 - July 2022 (v1.0.0) [Data set]. Zenodo. link

  • Buscombe, D., et al. (2022) Yearly CDIP:MOP-alongshore modeled wave statistics for California, January 2000 - July 2022 (v1.0.0) [Data set]. Zenodo. link

  • Buscombe, D., et al. (2022) Shoreline data at 30-m spatial resolution for 2001 coastal provinces or regions of the world, in geoJSON format. (v1.0.0) [Data set]. Zenodo. link

  • Buscombe, D., et al. (2022) Shoreline data at 30-m spatial resolution for 298 coastal counties of the conterminous USA, in geoJSON format. (v0.0.1) [Data set]. Zenodo. link

  • Buscombe, D., et al. (2022) Preliminary Coastal Grain Size Portal (C-GRASP) dataset. Version 1, January 2022, Zenodo link

  • Goldstein et al. (2022) Labels for Hurricane Florence (2018) Emergency Response Imagery from NOAA. 10.6084/m9.figshare.11604192.v1 link

  • Goldstein et al. (2022) Labels for Emergency Response Imagery from Hurricane Florence, Hurricane Michael, and Hurricane Isaias (Version 1.0) link

Coast Train

  • Wernette, P., et al. (2022) Coast Train--Labeled imagery for training and evaluation of data-driven models for image segmentation: U.S. Geological Survey data release, link. Check out the website for more details.

Other USGS ScienceBase data releases

  • Buscombe, D., Lundine, M.A., Janda, C.N., & Batiste, S. (2025, in review) Labeled satellite imagery for training machine learning models that predict the suitability of semantic segmentation model outputs for shoreline extraction. U.S. Geological Survey Data Release 10.5066/P1N4VI7H link

  • Buscombe, D., Lundine, M.A., Janda, C.N., & Batiste, S. (2025, in review) Labeled satellite imagery for training machine learning models that predict the suitability of imagery for shoreline extraction. U.S. Geological Survey Data Release 10.5066/P14MDKVJ link

  • Buscombe, D. et al. [2025, in prep] Satellite-derived shorelines for the U.S. states of Oregon and Washington for the period 1984-2023, obtained using CoastSat, U.S. Geological Survey data release.

  • Buscombe, D. et al. (2025) Satellite-derived shorelines from CoastSeg in multiple U.S. locations (1984-2023): U.S. Geological Survey data release, https://doi.org/10.5066/P1NUEFDP. link

  • Warrick, J.A., Buscombe, D. (2025) Data to Support Analyses of Shoreline Seasonal Cycles for Beaches of California: U.S. Geological Survey data release, https://doi.org/10.5066/P14WWHOJ. link

  • Janda, C.N. et al. (2025) Shoreline Change of Western Long Island, New York from Satellite Derived Shorelines: U.S. Geological Survey data release, https://doi.org/10.5066/P14VFVGZ. link

  • Buscombe, D. et al. (2024) Satellite-derived shorelines for the U.S. Gulf Coast states of Texas, Louisiana, Mississippi, and Florida for the period 1984-2022, obtained using CoastSat, U.S. Geological Survey data release, https://doi.org/10.5066/P1WFZXDM. link

  • Ritchie, A.C. et al. (2023) Remote Sensing Coastal Change Simple Data Distribution Service: U.S. Geological Survey data service, https://doi.org/10.5066/P9M3NYWI link

  • Ritchie, A.C. et al. (2022) Unprocessed aerial imagery from 9 December 2015 coastal survey of Central California.: U.S. Geological Survey data release, https://doi.org/10.5066/P9M3NYWI. link

  • Ritchie, M., et al. (2022) Aerial photogrammetry data and products of the North Carolina coast: U.S. Geological Survey data release. link

  • Kaplinski, M., et al. (2022) Channel mapping Glen Canyon Dam to Lees Ferry in Glen Canyon National Recreation Area, Arizona - Data: U.S. Geological Survey data release, link

  • Dean, D.J., et al.(2022) Suspended-sediment, bedload, bed-sediment, and multibeam sonar data in the Chippewa River, WI: U.S. Geological Survey data release, link

  • Ritchie, A.C. et al. (2021) Aerial photogrammetry data and products of the North Carolina coast—2018-10-06 to 2018-10-08, post-Hurricane Florence: U.S Geological Survey data release, link

  • Kranenburg et al. (2020) Post-Hurricane Florence aerial imagery: Cape Fear to Duck, North Carolina, October 6–8, 2018: U.S. Geological Survey data release. link

  • Kasprak et al. (2018) River Valley Sediment Connectivity Data, Colorado River, Grand Canyon, Arizona: U.S. Geological Survey data release. link

Other data releases

  • Buscombe, D., Goldstein, E. G., Sherwood, C. R., Bodine, C., Favela, J., Fitzpatrick, S., et al. (2022). Dataset accompanying Buscombe et al. Human-in-the-loop segmentation of Earth surface imagery. link

Demo Apps

Hugging Face

Streamlit

Pinned Loading

  1. DigitalGrainSize/SediNet DigitalGrainSize/SediNet Public

    Deep learning framework for optical granulometry (estimation of sedimentological variables from sediment imagery)

    Python 78 16

  2. LearnImageSegmentationWithUnets/UNets4LandscapeClassification LearnImageSegmentationWithUnets/UNets4LandscapeClassification Public

    Teaching resources for landscape/use/form classification with U-Nets

    Jupyter Notebook 3

  3. Doodleverse/dash_doodler Doodleverse/dash_doodler Public

    Doodler. A web application built with plotly/dash for image segmentation with minimal supervision. Plays nicely with segmentation gym, https://github.com/Doodleverse/segmentation_gym

    Python 67 13