Skip to content

6_Satellite Models

Jump to bottom
Daniel Buscombe edited this page Feb 23, 2023 · 1 revision

These models have been made for the CoastSeg package for Landsat and Sentinel-2 imagery. Imagery must be in 8-bit jpeg format.

💾 2-class

  • A dataset of high resolution (cm - m) oblique aerial imagery, labeled according to the following classes:
  1. Water
  2. Other

Example labeled imagery

formodelnoaug_ex16 formodelnoaug_ex23 formodelnoaug_ex47 formodelnoaug_ex64 formodelnoaug_ex71

Model for 512x512 RGB imagery

  • Zenodo release: https://zenodo.org/record/7448405
  • model citation: 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. https://doi.org/10.5281/zenodo.7448405
  • segmentation zoo model name: sat_RGB_2class_7448405

Training curve for the best model:

Some example validation outputs:

sat2class_rgb_512_v5_val_315 sat2class_rgb_512_v5_val_316 sat2class_rgb_512_v5_val_317 sat2class_rgb_512_v5_val_318 sat2class_rgb_512_v5_val_319

Model for 512x512 RGB+NIR+SWIR imagery

  • Zenodo release: https://zenodo.org/record/7448390
  • model citation: 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. https://doi.org/10.5281/zenodo.7448390
  • segmentation zoo model name: sat_5band_2class_7448390

Training curve for the best model:

Some example validation outputs:

sat2class_5d_512_v4_val_635 sat2class_5d_512_v4_val_636 sat2class_5d_512_v4_val_637 sat2class_5d_512_v4_val_638 sat2class_5d_512_v4_val_639

Model for 512x512 NDWI imagery

  • Zenodo release: https://zenodo.org/record/7557072
  • model citation: 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. https://doi.org/10.5281/zenodo.7557072
  • segmentation zoo model name: sat_NDWI_2class_7557072

Training curve for the best model:

Some example validation outputs:

sat2class_ndwi_512_v5_val_795 sat2class_ndwi_512_v5_val_796 sat2class_ndwi_512_v5_val_797 sat2class_ndwi_512_v5_val_798 sat2class_ndwi_512_v5_val_799

Model for 512x512 MNDWI imagery

  • Zenodo release: https://zenodo.org/record/7557080
  • model citation: 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. https://doi.org/10.5281/zenodo.7557080
  • segmentation zoo model name: sat_MNDWI_2class_7557080

Training curve for the best model:

Some example validation outputs:

sat2class_mndwi_512_v5_val_795 sat2class_mndwi_512_v5_val_796 sat2class_mndwi_512_v5_val_797 sat2class_mndwi_512_v5_val_798 sat2class_mndwi_512_v5_val_799

💾 4-class

  • A dataset of high resolution (cm - m) oblique aerial imagery, labeled according to the following classes:
  1. Water
  2. Whitewater
  3. Sediment
  4. Other
  • dataset: https://zenodo.org/record/7335647#.Y_fk6IDMLRY
  • dataset citation: Buscombe, Daniel, Goldstein, Evan, Bernier, Julie, Bosse, Stephen, Colacicco, Rosa, Corak, Nick, Fitzpatrick, Sharon, del Jesús González Guillén, Anais, Ku, Venus, Paprocki, Julie, Platt, Lindsay, Steele, Bethel, Wright, Kyle, & Yasin, Brandon. (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. https://doi.org/10.5281/zenodo.7335647

Example labeled imagery

formodelnoaug_ex3 formodelnoaug_ex11 formodelnoaug_ex15 formodelnoaug_ex17 formodelnoaug_ex79

Model for 512x512 RGB imagery

  • Zenodo release: https://zenodo.org/record/6950472
  • model citation: Buscombe, Daniel. (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. https://doi.org/10.5281/zenodo.6950472
  • segmentation zoo model name: sat_RGB_4class_6950472

Training curve for the best model:

Some example validation outputs:

sat4class_rgb_512_v1_val_635 sat4class_rgb_512_v1_val_636 sat4class_rgb_512_v1_val_637 sat4class_rgb_512_v1_val_638 sat4class_rgb_512_v1_val_639

Model for 512x512 RGB+NIR+SWIR imagery

  • Zenodo release: https://zenodo.org/record/7344606
  • model citation: Buscombe, Daniel. (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. https://doi.org/10.5281/zenodo.7344606
  • segmentation zoo model name: sat_5band_4class_7344606

Training curve for the best model:

Some example validation outputs:

sat4class_5d_512_v1_val_635 sat4class_5d_512_v1_val_636 sat4class_5d_512_v1_val_637 sat4class_5d_512_v1_val_638 sat4class_5d_512_v1_val_639

Model for 512x512 R+G+B+NIR+SWIR+NDWI+MNDW imagery

  • Zenodo release: https://zenodo.org/record/7358284
  • model citation: Buscombe, Daniel. (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. https://doi.org/10.5281/zenodo.7358284
  • segmentation zoo model name: sat_7band_4class_7358284

Training curve for the best model:

Some example validation outputs:

sat4class_7d_512_v1_val_635 sat4class_7d_512_v1_val_636 sat4class_7d_512_v1_val_637 sat4class_7d_512_v1_val_638 sat4class_7d_512_v1_val_639

Model for 512x512 NDWI imagery

  • Zenodo release: https://zenodo.org/record/7352859
  • model citation: Buscombe, Daniel. (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. https://doi.org/10.5281/zenodo.7352859
  • segmentation zoo model name: sat_NDWI_4class_7352859

Training curve for the best model:

Some example validation outputs:

sat4class_ndwi_512_v1_val_635 sat4class_ndwi_512_v1_val_636 sat4class_ndwi_512_v1_val_637 sat4class_ndwi_512_v1_val_638 sat4class_ndwi_512_v1_val_639

Model for 512x512 MNDWI imagery

  • Zenodo release: https://zenodo.org/record/7352850
  • model citation: Buscombe, Daniel. (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. https://doi.org/10.5281/zenodo.7352850
  • segmentation zoo model name: sat_MNDWI_4class_7352850

Training curve for the best model:

sat4class_mndwi_512_v1_trainhist_32

Some example validation outputs:

sat4class_mndwi_512_v1_val_635 sat4class_mndwi_512_v1_val_636 sat4class_mndwi_512_v1_val_637 sat4class_mndwi_512_v1_val_638 sat4class_mndwi_512_v1_val_639

Clone this wiki locally