5_Orthomosaic Models

These models have been made for the Seg2Map package for application of <=1m nadir and orthomosaic imagery

💾 enviroatlas

• EnviroAtlas is a dataset for generic landuse/cover segmentation of high-resolution orthomosaic/satellite imagery. Classes:

0. nodata
1. water
2. impervious
3. barren
4. trees
5. herbaceous
6. shrubland

• EnviroAtlas dataset

Example labeled imagery

Models for 6-class / 512x512 imagery

• Zenodo release: https://zenodo.org/record/7576909#.Y-FvcxzMLRY
• 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
• segmentation zoo model name: enviroatlas_6class_7576909

Training curve for the best model:

Some example validation outputs:

💾 open earth map

• OpenEarthNet is a dataset for generic landuse/cover segmentation of high-resolution orthomosaic imagery. Classes:

0. bareland
1. rangeland
2. development
3. road
4. tree
5. water
6. agricultural
7. building
8. nodata

• OpenEarthMap
• 9 class dataset (bareland, rangeland, dev., road, tree, water, ag., building, nodata)

Example labeled imagery

Model for 9-classes / 512x512 imagery

• Zenodo release: https://zenodo.org/record/7576894#.Y-FsiRzMLRY
• 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
• segmentation zoo model name: openearthmap_9class_7576894

Training curve for the best model:

Some example validation outputs:

💾 deepglobe

• DeepGlobe is a dataset for generic landuse/cover segmentation of high-resolution orthomosaic/satellite imagery. Classes:

0. urban
1. agricultural
2. rangeland
3. forest
4. water
5. bare
6. unknown

• DeepGlobe / aerial / high-res. sat
• 7 class dataset (urban, ag., rangeland, forest, water, bare, unknown)

Example labeled imagery

512x512 imagery

• Zenodo release: https://zenodo.org/record/7576898#.Y-FtUBzMLRY
• 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
• segmentation zoo model name: deepglobe_7class_7576898

Training curve for the best model:

Some example validation outputs:

💾 AAAI-buildings

• AAAI-Buildings is a dataset for segmentation of buildings in hurricane-affected areas in high-resolution imagery. A dataset of high resolution (m) orthorectified aerial imagery, labeled according to the following classes:

1. Other
2. Building
3. Flooded Building

• dataset: https://github.com/FrontierDevelopmentLab/multi3net
• dataset citation: Rudner, T. G. J.; Rußwurm, M.; Fil, J.; Pelich, R.; Bischke, B.; Kopačková, V.; Biliński, P. Segmenting Flooded Buildings via Fusion of Multiresolution, Multisensor, and Multitemporal Satellite Imagery. In AAAI 2019. https://arxiv.org/pdf/1812.01756.pdf

Example labeled imagery

Model for 1024x1024 imagery, building/no-building

• Zenodo release: https://doi.org/10.5281/zenodo.7607895
• model citation: Doodleverse/Segmentation Zoo/Seg2Map Res-UNet models for segmentation of buildings of RGB 1024x1024 high-res. images (v1.0) [Data set]. Zenodo. https://doi.org/10.5281/zenodo.7607895
• segmentation zoo model name: aaai_building_7607895
• Classes:

1. Other
2. Building

Training curve for the best model:

Some example validation outputs:

Model for 1024x1024 imagery, flooded-building/other

• Zenodo release: https://doi.org/10.5281/zenodo.7622733
• model citation: Buscombe, Daniel. (2023). Doodleverse/Segmentation Zoo/Seg2Map Res-UNet models for segmentation of AAAI/flooded buildings in RGB 1024x1024 high-res. images (v1.0) [Data set]. Zenodo. https://doi.org/10.5281/zenodo.7622733
• segmentation zoo model name: aaai_floodedbuildings_7622733
• Classes:

1. Other
2. Flooded Building

Training curve for the best model:

Some example validation outputs:

💾 XBD

• XBD-hurricanes subset is a dataset for segmentation of buildings in hurricane-affected areas in high-resolution imagery. A dataset of high resolution (cm - m) oblique aerial imagery, labeled according to the following classes:

1. No-damage
2. Minor-damage
3. Major-damage
4. No-building

• dataset: https://arxiv.org/abs/1911.09296
• dataset citation: Ritwik Gupta, Bryce Goodman, Nirav Patel, Ricky Hosfelt, Sandra Sajeev, Eric Heim, Jigar Doshi, Keane Lucas, Howie Choset, and Matthew Gaston. Creating xbd: A dataset for assessing building damage from satellite imagery. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops, June 2019. https://arxiv.org/abs/1911.09296

Example labeled imagery

Model for 768x768 imagery, building/no-building

• Zenodo release: https://zenodo.org/record/7613212
• model citation: Buscombe, Daniel. (2023). Doodleverse/Segmentation Zoo/Seg2Map SegFormer models for segmentation of xBD/buildings in RGB 768x768 high-res. images (v1.0) [Data set]. Zenodo. https://doi.org/10.5281/zenodo.7613212
• segmentation zoo model name: xbd_building_7613212
• classes: 1. other, 2. building

Training curve for the best model:

Some example validation outputs:

Model for 768x768 imagery, building damage

• Zenodo release: https://doi.org/10.5281/zenodo.7613175
• model citation: Buscombe, Daniel. (2023). Doodleverse/Segmentation Zoo/Seg2Map SegFormer models for segmentation of xBD/damaged buildings in RGB 768x768 high-res. images (v1.0) [Data set]. Zenodo. https://doi.org/10.5281/zenodo.7613175
• segmentation zoo model name: xbd_damagedbuilding_7613175
• classes: 1. no-damage, 2. minor-damage, 3. major-damage, 4. no building

Training curve for the best model:

Some example validation outputs: