Dataset

This dataset was created by Jitendra Sharma

Contents

CT data and corresponding mask images used to train 3D-Unet

This data set contains: - Training dataset: 271 CT-scans of inner ears used for optimization and training of the model. - Validation dataset: 70 CT-scans of inner ears used for external validation. - U-net architecture deep-learning model's weight after optimized training. - All manual segmentations performed for both datasets. - All post-processed automated segmentations performed by the model for bothd atasets.

Download our Sentinel-2 imagery dataset to facilitate your model training processes. This high-resolution, multispectral collection is ideal for applications in emergency response, transportation, and wildfire detection. This training area covers the communities of Corona, Ventura, and Pala Mesa California.RS_Data_MetaData contains the exact subset of imagery used for our Wildfire Threat Detection for Transportation Infrastructure using U-Net for Semantic Segmentation notebook.

Doodleverse/Segmentation Zoo/Seg2Map Res-UNet models for CoastTrain water/other segmentation of RGB 768x768 orthomosaic images

These Residual-UNet model data are based on Coast Train images and associated labels. https://coasttrain.github.io/CoastTrain/docs/Version%201:%20March%202022/data

Models have been created using Segmentation Gym* using the following dataset**: https://doi.org/10.1038/s41597-023-01929-2

Image size used by model: 768 x 768 x 3 pixels

classes: 1. Water 2. Other

File descriptions

For each model, there are 5 files with the same root name:

1. '.json' config file: this is the file that was used by Segmentation Gym* to create the weights file. It contains instructions for how to make the model and the data it used, as well as instructions for how to use the model for prediction. It is a handy wee thing and mastering it means mastering the entire Doodleverse.

2. '.h5' weights file: this is the file that was created by the Segmentation Gym* function train_model.py. It contains the trained model's parameter weights. It can called by the Segmentation Gym* function seg_images_in_folder.py. Models may be ensembled.

3. '_modelcard.json' model card file: this is a json file containing fields that collectively describe the model origins, training choices, and dataset that the model is based upon. There is some redundancy between this file and the config file (described above) that contains the instructions for the model training and implementation. The model card file is not used by the program but is important metadata so it is important to keep with the other files that collectively make the model and is such is considered part of the model

4. '_model_history.npz' model training history file: this numpy archive file contains numpy arrays describing the training and validation losses and metrics. It is created by the Segmentation Gym function train_model.py

5. '.png' model training loss and mean IoU plot: this png file contains plots of training and validation losses and mean IoU scores during model training. A subset of data inside the .npz file. It is created by the Segmentation Gym function train_model.py

Additionally, BEST_MODEL.txt contains the name of the model with the best validation loss and mean IoU

References *Segmentation Gym: Buscombe, D., & Goldstein, E. B. (2022). A reproducible and reusable pipeline for segmentation of geoscientific imagery. Earth and Space Science, 9, e2022EA002332. https://doi.org/10.1029/2022EA002332 See: https://github.com/Doodleverse/segmentation_gym

**Buscombe, D., Wernette, P., Fitzpatrick, S. et al. A 1.2 Billion Pixel Human-Labeled Dataset for Data-Driven Classification of Coastal Environments. Sci Data 10, 46 (2023). https://doi.org/10.1038/s41597-023-01929-2

Dataset

This dataset was created by Awsaf

Contents

Doodleverse/Segmentation Zoo/Seg2Map Res-UNet models for FloodNet/10-class segmentation of RGB 768x512 UAV images

These Residual-UNet model data are based on FloodNet images and associated labels.

Models have been created using Segmentation Gym* using the following dataset**: https://github.com/BinaLab/FloodNet-Challenge-EARTHVISION2021

Image size used by model: 768 x 512 x 3 pixels

classes: 1. Background 2. Building-flooded 3. Building-non-flooded 4. Road-flooded 5. Road-non-flooded 6. Water 7. Tree 8. Vehicle 9. Pool 10. Grass

File descriptions

For each model, there are 5 files with the same root name:

1. '.json' config file: this is the file that was used by Segmentation Gym* to create the weights file. It contains instructions for how to make the model and the data it used, as well as instructions for how to use the model for prediction. It is a handy wee thing and mastering it means mastering the entire Doodleverse.

2. '.h5' weights file: this is the file that was created by the Segmentation Gym* function train_model.py. It contains the trained model's parameter weights. It can called by the Segmentation Gym* function seg_images_in_folder.py. Models may be ensembled.

3. '_modelcard.json' model card file: this is a json file containing fields that collectively describe the model origins, training choices, and dataset that the model is based upon. There is some redundancy between this file and the config file (described above) that contains the instructions for the model training and implementation. The model card file is not used by the program but is important metadata so it is important to keep with the other files that collectively make the model and is such is considered part of the model

4. '_model_history.npz' model training history file: this numpy archive file contains numpy arrays describing the training and validation losses and metrics. It is created by the Segmentation Gym function train_model.py

5. '.png' model training loss and mean IoU plot: this png file contains plots of training and validation losses and mean IoU scores during model training. A subset of data inside the .npz file. It is created by the Segmentation Gym function train_model.py

Additionally, BEST_MODEL.txt contains the name of the model with the best validation loss and mean IoU

images.zip and labels.zip contain the images and labels, respectively, used to train the model

References *Segmentation Gym: Buscombe, D., & Goldstein, E. B. (2022). A reproducible and reusable pipeline for segmentation of geoscientific imagery. Earth and Space Science, 9, e2022EA002332. https://doi.org/10.1029/2022EA002332 See: https://github.com/Doodleverse/segmentation_gym

** Rahnemoonfar, M., Chowdhury, T., Sarkar, A., Varshney, D., Yari, M. and Murphy, R.R., 2021. Floodnet: A high resolution aerial imagery dataset for post flood scene understanding. IEEE Access, 9, pp.89644-89654.

Fully annotated dataset for training of 3D segmentation models. We provide the Raw image patches in the Raw directory and instance segmentation labels in the RealMask directory, the semantic segmentation masks are provided in the BinaryMask directory. Manually curated from originally published dataset over here: http://celltrackingchallenge.net/3d-datasets/ by team at Kapoorlabs.

Doodleverse/Segmentation Zoo/Seg2Map Res-UNet models for OpenEarthMap/9-class segmentation of RGB 512x512 high-res. images

These Residual-UNet model data are based on the OpenEarthMap dataset

Models have been created using Segmentation Gym* using the following dataset**: https://zenodo.org/record/7223446#.Y9gtWHbMIuV

Image size used by model: 512 x 512 x 3 pixels

classes: 1. bareland 2. rangeland 3. development 4. road 5. tree 6. water 7. agricultural 8. building 9. nodata

File descriptions

For each model, there are 5 files with the same root name:

1. '.json' config file: this is the file that was used by Segmentation Gym* to create the weights file. It contains instructions for how to make the model and the data it used, as well as instructions for how to use the model for prediction. It is a handy wee thing and mastering it means mastering the entire Doodleverse.

2. '.h5' weights file: this is the file that was created by the Segmentation Gym* function train_model.py. It contains the trained model's parameter weights. It can called by the Segmentation Gym* function seg_images_in_folder.py. Models may be ensembled.

3. '_modelcard.json' model card file: this is a json file containing fields that collectively describe the model origins, training choices, and dataset that the model is based upon. There is some redundancy between this file and the config file (described above) that contains the instructions for the model training and implementation. The model card file is not used by the program but is important metadata so it is important to keep with the other files that collectively make the model and is such is considered part of the model

4. '_model_history.npz' model training history file: this numpy archive file contains numpy arrays describing the training and validation losses and metrics. It is created by the Segmentation Gym function train_model.py

5. '.png' model training loss and mean IoU plot: this png file contains plots of training and validation losses and mean IoU scores during model training. A subset of data inside the .npz file. It is created by the Segmentation Gym function train_model.py

Additionally, BEST_MODEL.txt contains the name of the model with the best validation loss and mean IoU

References *Segmentation Gym: Buscombe, D., & Goldstein, E. B. (2022). A reproducible and reusable pipeline for segmentation of geoscientific imagery. Earth and Space Science, 9, e2022EA002332. https://doi.org/10.1029/2022EA002332 See: https://github.com/Doodleverse/segmentation_gym

**Xia, Yokoya, Adriano, & Broni-Bediako. (2022). OpenEarthMap: A Benchmark Dataset for Global High-Resolution Land Cover Mapping [Data set]. Zenodo. https://doi.org/10.5281/zenodo.7223446

An R object (list) containing two matrices (tensors) corresponding the untiled brightfield images and their computed mask

Training and test images of live B. subtilis cells expressing FtsZ-GFP for the task of segmentation.

Additional information can be found on this github wiki.

The example shows the fluorescence widefield image of live B. subtilis cells expressing FtsZ-GFP, the manually annotated instance segmentation mask and the corresponding 2-label semantic segmentation mask used for model training.

Training and test dataset

Data type: Paired fluorescence and segmented mask images

Microscopy data type: 2D widefield images (fluorescence)

Microscope: Custom-built 100x inverted microscope bearing a 100x TIRF objective (Nikon CFI Apochromat TIRF 100XC Oil); images were captured on a Prime BSI sCMOS camera (Teledyne Photometrics)

Cell type: B. subtilis strain SH130 grown under agarose pads

File format: .tiff (8-bit)

Image size: 1024 x 1024 px² (Pixel size: 65 nm)

Image preprocessing: Images were denoised using PureDenoise and resulting 32-bit images were converted into 8-bit images after normalizing to 1% and 99.98% percentiles. Images were manually annotated using the Labkit Fiji plugin and mask images with labeled cytosol and cell boundaries were created using a custom Fiji macro (see our github repository).

Multi-label U-Net model:

The U-Net (2D) multilabel model was generated using the ZeroCostDL4Mic platform (Chamier & Laine et al., 2021). It was trained from scratch for 200 epochs on 733 paired image patches (image dimensions: (1024 x 1024 px²), patch size: (256 x 256 px²)) with a batch size of 8 and a categorical_crossentrop loss function, using the U-Net (2D) multilabel ZeroCostDL4Mic notebook (v 1) (Chamier & Laine et al., 2021). Key python packages used include tensorflow (v 0.1.12), Keras (v 2.3.1), numpy (v 1.19.5), cuda (v 11.1.105). The training was accelerated using a Tesla P100GPU.

Author(s): Mia Conduit1,2, Séamus Holden1,3

Contact email: Seamus.Holden@newcastle.ac.uk

Affiliation:

1) Centre for Bacterial Cell Biology, Biosciences Institute, Newcastle University, NE2 4AX UK

2) ORCID: 0000-0002-7169-907X

Associated publications: Whitley et al., 2021, Nature Communications, https://doi.org/10.15252/embj.201696235

In this research, the best management practices include vegetative/structural conservation practices (SCP) across crop fields, such as grassed waterways and terraces. This reference dataset includes 500,000 pair patches (false-color image (B1: NIR, B2: Red, B3: Green) and binary label (SCP: yes[1] or no[0]). These training samples were randomly extracted from Iowa BMP project (https://www.gis.iastate.edu/gisf/projects/conservation-practices) and present 90% of patches with SCP areas and 10% of patches non-SCP area. The patch dimension is 256 x 256 pixels at 2-m resolution. Due to the file size, the images were upload in different *.rar files (imagem_0_200k.rar, imagem_200_400k.rar, imagem_400_500k.rar), and the user should download all and merge them in the same folder. The corresponding labels are all in "class_bin.rar" file.

Application: These pair images are useful for conservation practitioners interested in the classification of vegetative/structural SCPs using deep-learning semantic segmentation methods.

Further information will be available in future.

Dataset

This dataset was created by Pushpa Pandey

Contents

This dataset visualizes Grad-CAM based heatmap applied on membrane segmentation results when using Unet. The training data are shown in "train" folder in which: - "checkpoint" folder: stores checkpoint files for 3 epochs: 100, 500, and 5,000. - "image" folder: holds training images - "label" folder: stores labelled membrane images The testing results are stored in "test_xxx" folders for 3 epochs: 100, 500 and 5,000.

Training data for line segmentation of Tibetan woodblock prints. This version of dataset contain images randomly picked images from LOC scans of the derge kangyur.

This dataset was trained with Unet Model using fastai library. The result was unsatisfactory due to insufficient data.

Datasets contains:

Images/ : Original pecha image

Labels/ : Mask of each image

valid.txt: list of validation filename

code.txt: list of name of object

This Residual-UNet model is trained on 1,179 pairs of human-generated segmentation labels and images from Emergency Response Imagery (ERI) collected by US National Oceanic and Atmospheric Administration (NOAA) after Hurricane Barry, Delta, Dorian, Florence, Ida, Laura, Michael, Sally, Zeta, and Tropical Storm Gordon. The dataset is available here: https://doi.org/10.5281/zenodo.7268082

Models have been created using Segmentation Gym:

Code - https://github.com/Doodleverse/segmentation_gym

Paper - https://doi.org/10.1029/2022EA002332

The model takes input images that are 512 x 512 x 3 pixels, and the output is 512 x 512 x 4, corresponding to 4 classes:

water

bare sediment

vegetation

development (roads, buildings, power lines, parking lots, etc.)

Included here are 6 files with the same root name:

'.json' config file: this is the file that was used by Segmentation Gym to create the weights file. It contains instructions for how to make the model and the data it used, as well as instructions for how to use the model for prediction.

'.h5' weights file: this is the file that was created by the Segmentation Gym function train_model.py. It contains the trained model's parameter weights. It can called by the Segmentation Gym function seg_images_in_folder.py.

'_model_history.npz' model training history file: this numpy archive file contains numpy arrays describing the training and validation losses and metrics. It is created by the Segmentation Gym function train_model.py

'.png' model training loss and mean IoU plot: this png file contains plots of training and validation losses and mean IoU scores during model training. A subset of data inside the .npz file. It is created by the Segmentation Gym function train_model.py

'.zip' of the model in the Tensorflow ‘saved model’ format. It is created by the Segmentation Gym function utils/gen_saved_model.py

'_modelcard.json' model card file: this is a json file containing fields that collectively describe the model origins, training choices, and dataset that the model is based upon. There is some redundancy between this file and the config file (described above) that contains the instructions for the model training and implementation. The model card file is not used by the program but is important metadata so it is important to keep with the other files that collectively make the model and is such is considered part of the model

Additionally, BEST_MODEL.txt contains the name of the model with the best validation loss and mean IoU

Doodleverse/Segmentation Gym Res-UNet models for 2-class (water, other) segmentation of CoastCam runup timestack imagery

This model release is part of the Doodleverse: https://github.com/Doodleverse

These Residual-UNet model data are based on RGB (red, green, and blue) images of coasts and associated labels.

Models have been created using Segmentation Gym* using an as-yet unpublished dataset of images and associated label images. See https://github.com/Doodleverse for more information about how this model was trained, and how to use it for inference

Classes: {0=other, 1=water}

File descriptions

There are two models; v7 has been trained from scratch, and v8 has been fine-tuned using hyperparameter adjustment. For each model, there are 5 files with the same root name:

1. '.json' config file: this is the file that was used by Segmentation Gym* to create the weights file. It contains instructions for how to make the model and the data it used, as well as instructions for how to use the model for prediction. It is a handy wee thing and mastering it means mastering the entire Doodleverse.

2. '.h5' weights file: this is the file that was created by the Segmentation Gym* function train_model.py. It contains the trained model's parameter weights. It can called by the Segmentation Gym* function seg_images_in_folder.py. Models may be ensembled.

3. '_modelcard.json' model card file: this is a json file containing fields that collectively describe the model origins, training choices, and dataset that the model is based upon. There is some redundancy between this file and the config file (described above) that contains the instructions for the model training and implementation. The model card file is not used by the program but is important metadata so it is important to keep with the other files that collectively make the model and is such is considered part of the model

4. '_model_history.npz' model training history file: this numpy archive file contains numpy arrays describing the training and validation losses and metrics. It is created by the Segmentation Gym function train_model.py

5. '.png' model training loss and mean IoU plot: this png file contains plots of training and validation losses and mean IoU scores during model training. A subset of data inside the .npz file. It is created by the Segmentation Gym function train_model.py

Additionally,

1. BEST_MODEL.txt contains the name of the model with the best validation loss and mean IoU
2. sample_images.zip contains a few example input files, for model testing

References

*Segmentation Gym: Buscombe, D., & Goldstein, E. B. (2022). A reproducible and reusable pipeline for segmentation of geoscientific imagery. Earth and Space Science, 9, e2022EA002332. https://doi.org/10.1029/2022EA002332 See: https://github.com/Doodleverse/segmentation_gym

The model was trained on over 3,000 image-mask pairs using Sentinel-2 imagery and CAL FIRE’s Wildland Fire Threat layer, specifically from the 2017 Lilac Fire in San Diego County, California. The training data was curated from the Pala Mesa region, an area significantly impacted by wildfire activity and containing critical road and rail networks.The model outputs multi-class segmentation masks that classify areas into low-moderate, high, very high, and extreme wildfire threat categories which may support emergency preparedness and response and infrastructure risk analysis. This pretrained version allows users to test the model and generate inferences without requiring a long runtime for training, making it ideal for rapid evaluation, demonstration, or integration into spatial AI workflows.NoteThis model was trained on a region-specific dataset (Lilac Fire, 2017), and generalization to other fires or regions may require fine-tuning.Masks are aligned with CAL FIRE’s threat classification and were processed using 1,000-foot buffers around infrastructure features.Data provided by:California Department of Forestry and Fire Protection (CAL FIRE)European Space Agency (Sentinel-2 via Copernicus Program)U.S. Census Bureau (TIGER/Line Roads)Federal Railroad Administration / Bureau of Transportation Statistics

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.

These Residual-UNet model data are based on RGB (red, green, and blue) images of coasts and associated labels.

Models have been created using Segmentation Gym* using the following dataset**: https://doi.org/10.5281/zenodo.7335647

Classes: {0=water, 1=whitewater, 2=sediment, 3=other}

File descriptions

For each model, there are 5 files with the same root name:

1. '.json' config file: this is the file that was used by Segmentation Gym* to create the weights file. It contains instructions for how to make the model and the data it used, as well as instructions for how to use the model for prediction. It is a handy wee thing and mastering it means mastering the entire Doodleverse.

2. '.h5' weights file: this is the file that was created by the Segmentation Gym* function `train_model.py`. It contains the trained model's parameter weights. It can called by the Segmentation Gym* function `seg_images_in_folder.py`. Models may be ensembled.

3. '_modelcard.json' model card file: this is a json file containing fields that collectively describe the model origins, training choices, and dataset that the model is based upon. There is some redundancy between this file and the `config` file (described above) that contains the instructions for the model training and implementation. The model card file is not used by the program but is important metadata so it is important to keep with the other files that collectively make the model and is such is considered part of the model

4. '_model_history.npz' model training history file: this numpy archive file contains numpy arrays describing the training and validation losses and metrics. It is created by the Segmentation Gym function `train_model.py`

5. '.png' model training loss and mean IoU plot: this png file contains plots of training and validation losses and mean IoU scores during model training. A subset of data inside the .npz file. It is created by the Segmentation Gym function `train_model.py`

Additionally, BEST_MODEL.txt contains the name of the model with the best validation loss and mean IoU

References

*Segmentation Gym: Buscombe, D., & Goldstein, E. B. (2022). A reproducible and reusable pipeline for segmentation of geoscientific imagery. Earth and Space Science, 9, e2022EA002332. https://doi.org/10.1029/2022EA002332 See: https://github.com/Doodleverse/segmentation_gym

** 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

The GIS database used in this project serves as a centralized repository for all spatial datasets required for wildfire threat analysis and model training. It includes CAL FIRE’s Wildland Fire Threat layer, which provides pixel-based classifications of wildfire potential across California, as well as transportation infrastructure layers, including primary and secondary roads and railways.To support impact analysis, 1,000-foot buffer zones were generated around each infrastructure feature to define zones of interest for wildfire segmentation. The database is structured for integration into both machine learning workflows and GIS environments, enabling seamless overlay, visualization, and spatial querying within platforms such as ArcGIS Pro or QGIS.