Rock Segmentation

## Overview

Rock Segmentation is a dataset for instance segmentation tasks - it contains Rocks annotations for 299 images.

## Getting Started

You can download this dataset for use within your own projects, or fork it into a workspace on Roboflow to create your own model.

  ## License

  This dataset is available under the [CC BY 4.0 license](https://creativecommons.org/licenses/CC BY 4.0).

Farmland Instance Segmentation

## Overview

Farmland Instance Segmentation is a dataset for instance segmentation tasks - it contains Test ZTTD annotations for 984 images.

## Getting Started

You can download this dataset for use within your own projects, or fork it into a workspace on Roboflow to create your own model.

  ## License

  This dataset is available under the [CC BY 4.0 license](https://creativecommons.org/licenses/CC BY 4.0).

Extra Segmentation Dataset

## Overview

Extra Segmentation Dataset is a dataset for object detection tasks - it contains Tomato annotations for 272 images.

## Getting Started

You can download this dataset for use within your own projects, or fork it into a workspace on Roboflow to create your own model.

  ## License

  This dataset is available under the [CC BY 4.0 license](https://creativecommons.org/licenses/CC BY 4.0).

Waste disposal is a global challenge, especially in densely populated areas. Efficient waste segregation is critical for separating recyclable from non-recyclable materials. While developed countries have established and refined effective waste segmentation and recycling systems, our country still uses manual segregation to identify and process recyclable items. This study presents a dataset intended to improve automatic waste segmentation systems. The dataset consists of 784 images that have been manually annotated for waste classification. These images were primarily taken in and around Jadavpur University, including streets, parks, and lawns. Annotations were created with the Labelme program and are available in color annotation formats. The dataset includes 14 waste categories: plastic containers, plastic bottles, thermocol, metal bottles, plastic cardboard, glass, thermocol plates, plastic, paper, plastic cups, paper cups, aluminum foil, cloth, and nylon. The dataset includes a total of 2350 object segments.Other Information:Published in: Mendely DataLicense: http://creativecommons.org/licenses/by/4.0/See dataset on publisher's website: https://data.mendeley.com/datasets/gr99ny6b8p/1

ATCS is a dataset designed to train deep learning models to volumetrically segment clouds from multi-angle satellite imagery. The dataset consists of spatiotemporally aligned patches of multi-angle polarimetry from the POLDER sensor aboard the PARASOL mission and vertical cloud profiles from the 2B-CLDCLASS product using the cloud profiling radar (CPR) aboard CloudSat.

"Image.zip" contains 955 corrrosion images, 1480 crack images, 1269 free lime images, 873 water leakage images, and 1244 spalling images. These images are labeled with numbers from 0 to 6 including the background. The "Label.zip" file contains the labeled images, and the "Image.json" file contains the label information.

Chest X-ray(CXR) images are prominent among medical images and are commonly administered in emergency diagnosis and treatment corresponding to cardiac and respiratory diseases. Though there are robust solutions available for medical diagnosis, validation of artificial intelligence (AI) in radiology is still questionable. Segmentation is pivotal in chest radiographs that aid in improvising the existing AI-based medical diagnosis process. We provide the CXLSeg dataset: Chest X-ray with Lung Segmentation, a comparatively large dataset of segmented Chest X-ray radiographs based on the MIMIC-CXR dataset, a popular CXR image dataset. The dataset contains segmentation results of 243,324 frontal view images of the MIMIC-CXR dataset and corresponding masks. Additionally, this dataset can be utilized for computer vision-related deep learning tasks such as medical image classification, semantic segmentation and medical report generation. Models using segmented images yield better results since only the features related to the important areas of the image are focused. Thus images of this dataset can be manipulated to any visual feature extraction process associated with the original MIMIC-CXR dataset and enhance the results of the published or novel investigations. Furthermore, masks provided by this dataset can be used to train segmentation models when combined with the MIMIC-CXR-JPG dataset. The SA-UNet model achieved a 96.80% in dice similarity coefficient and 91.97% in IoU for lung segmentation using CXLSeg.

Cov Khoom Siv Feem Ntau Segmentation Dataset ua haujlwm rau kev lag luam e-lag luam thiab kev lom zem pom kev lag luam nrog ntau cov duab sau hauv internet, muaj cov kev daws teeb meem xws li 800 × 600 txog 4160 × 3120. Cov ntaub ntawv no suav nrog ntau qhov sib txawv ntawm cov xwm txheej niaj hnub thiab cov khoom, suav nrog ntau tus neeg, tsiaj txhu thiab cov rooj tog zaum. segmentation.

Dataset Labels

['dry_joint', 'incorrect_installation', 'pcb_damage', 'short_circuit']

  Number of Images

{'valid': 25, 'train': 128, 'test': 36}

  How to Use

Install datasets:

pip install datasets

Load the dataset:

from datasets import load_dataset

ds = load_dataset("keremberke/pcb-defect-segmentation", name="full") example = ds['train'][0]

  Roboflow Dataset Page

https://universe.roboflow.com/diplom-qz7q6/defects-2q87r/dataset/8… See the full description on the dataset page: https://huggingface.co/datasets/keremberke/pcb-defect-segmentation.

This dataset provides comprehensive customer data suitable for segmentation analysis. It includes anonymized demographic, transactional, and behavioral attributes, allowing for detailed exploration of customer segments. Leveraging this dataset, marketers, data scientists, and business analysts can uncover valuable insights to optimize targeted marketing strategies and enhance customer engagement. Whether you're looking to understand customer behavior or improve campaign effectiveness, this dataset offers a rich resource for actionable insights and informed decision-making.

Key Features:

Anonymized demographic, transactional, and behavioral data. Suitable for customer segmentation analysis. Opportunities to optimize targeted marketing strategies. Valuable insights for improving campaign effectiveness. Ideal for marketers, data scientists, and business analysts.

Usage Examples:

Segmenting customers based on demographic attributes. Analyzing purchase behavior to identify high-value customer segments. Optimizing marketing campaigns for targeted engagement. Understanding customer preferences and tailoring product offerings accordingly. Evaluating the effectiveness of marketing strategies and iterating for improvement. Explore this dataset to unlock actionable insights and drive success in your marketing initiatives!

Delve into the Pupils Segmentation Dataset Essential for ophthalmology tech, AI driven vision studies, and advanced eye research.

The HaN-Seg: Head and Neck Organ-at-Risk CT & MR Segmentation Dataset is a publicly available dataset of anonymized head and neck (HaN) images of 42 patients that underwent both CT and T1-weighted MR imaging for the purpose of image-guided radiotherapy planning. In addition, the dataset also contains reference segmentations of 30 organs-at-risk (OARs) for CT images in the form of binary segmentation masks, which were obtained by curating manual pixel-wise expert image annotations. A full description of the HaN-Seg dataset can be found in:

G. Podobnik, P. Strojan, P. Peterlin, B. Ibragimov, T. Vrtovec, "HaN-Seg: The head and neck organ-at-risk CT & MR segmentation dataset", Medical Physics, 2023. 10.1002/mp.16197,

and any research originating from its usage is required to cite this paper.

In parallel with the release of the dataset, the HaN-Seg: The Head and Neck Organ-at-Risk CT & MR Segmentation Challenge is launched to promote the development of new and application of existing state-of-the-art fully automated techniques for OAR segmentation in the HaN region from CT images that exploit the information of multiple imaging modalities, in this case from CT and MR images. The task of the HaN-Seg challenge is to automatically segment up to 30 OARs in the HaN region from CT images in the devised test set, consisting of 14 CT and MR images of the same patients, given the availability of the training set (i.e. the herein publicly available HaN-Seg dataset), consisting of 42 CT and MR images of the same patients with reference 3D OAR binary segmentation masks for CT images.

Please find below a list of relevant publications that address: (1) the assessment of inter-observer and inter-modality variability in OAR contouring, (2) results of the HaN-Seg challenge, (3) development of our multimodal segmentation model, and (4) development of MR-to-CT image-to-image translation using diffusion models:

1. G. Podobnik, B. Ibragimov, P. Strojan, P. Peterlin, T. Vrtovec, "vOARiability: Interobserver and intermodality variability analysis in OAR contouring from head and neck CT and MR images", Medical Physics, 2024. 10.1002/mp.16924,
2. G. Podobnik, B. Ibragimov, E. Tappeiner, et al., "HaN-Seg: The head and neck organ-at-risk CT and MR segmentation challenge", Radiotherapy and Oncology, 2024. 10.1016/j.radonc.2024.110410,
3. G. Podobnik, P. Strojan, P. Peterlin, B. Ibragimov, T. Vrtovec, "Multimodal CT and MR Segmentation of Head and Neck Organs-at-Risk", MICCAI 2023, 2023. 10.1007/978-3-031-43901-8_71,
4. R. M. Šter, G. Podobnik, T. Vrtovec, "Diffusion-based MR-to-CT translation of head and neck images", SPIE Medical Imaging 2025, 2025. 10.1117/12.3047458.

The Liver Tumor Segmentation Benchmark (LiTS) dataset contains 130 CT scans of patients with liver cancer. This dataset includes 2D slices from 3D CT scans with masks for liver, tumor, bone, arteries, and kidneys.

This dataset facilitates slice based segmentation, which produces more accurate results (in most cases) than 3D segmentation.

Reference: https://doi.org/10.1016/j.media.2022.102680

File Description

This dataset contains the slices from the LiTS dataset in the format: Volume-{VolumeNumber}-{SliceNumber.png}.

Both the image and the mask files have the same naming convention.

We established a large-scale plant disease segmentation dataset named PlantSeg. PlantSeg comprises more than 11,400 images of 115 different plant diseases from various environments, each annotated with its corresponding segmentation label for diseased parts. To the best of our knowledge, PlantSeg is the largest plant disease segmentation dataset containing in-the-wild images. Our dataset enables researchers to evaluate their models and provides a valid foundation for the development and benchmarking of plant disease segmentation algorithms.

Please note that due to the image limitations of Roboflow, the dataset provided here is not complete.

Project page: https://github.com/tqwei05/PlantSeg

Paper: https://arxiv.org/abs/2409.04038

Complete dataset download: https://zenodo.org/records/13958858

Reference: @article{wei2024plantseg, title={PlantSeg: A Large-Scale In-the-wild Dataset for Plant Disease Segmentation}, author={Wei, Tianqi and Chen, Zhi and Yu, Xin and Chapman, Scott and Melloy, Paul and Huang, Zi}, journal={arXiv preprint arXiv:2409.04038}, year={2024} }

Mixed training and test images of S. aureus, E. coli and B. subtilis for cell segmentation using StarDist, as well as the trained StarDist model.

Additional information can be found on this github wiki.

Data type: Paired bright field / fluorescence and segmented mask images

Microscopy data type: 2D widefield images; DIC and fluorescence for S. aureus, bright field images for E. coli, and fluorescence images for B. subtilis

Microscopes:

S. aureus:

GE HealthCare Deltavision OMX system (with temperature and humidity control, 37°C) equipped with an Olympus 60x 1.42NA Oil immersion objective and 2 PCO Edge 5.5 sCMOS cameras (one for DIC, one for fluorescence)

E.coli:

Nikon Eclipse Ti-E equipped with an Apo TIRF 1.49NA 100x oil immersion objective

B. subtilis:

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 types: S. aureus strain JE2, E. coli MG1655 (CGSC #6300) and B. subtilis strain SH130; all grown under agarose pads

File format: .tif (8-bit and 16-bit)

Image size: 512 x 512 px² @ 80 nm pixel size (S. aureus); 1024 x 1024 px² @ 79 nm pixel size (E. coli); 1024 x 1024 px² @ 65 nm pixel size (B. subtilis)

Image preprocessing:

S. aureus:

Raw images were manually annotated by drawing ellipses in the NR fluorescence image and segmented images were created using the LOCI plugin (“ROI Map”). For training, images and masks were quartered into four 256 x 256 px² patches.

E. coli:

Raw images were recorded in 16-bit mode (image size 512x512 px² @ 158 nm/px). Images were upscaled with a factor of 2 (no interpolation) to enable generation of higher-quality segmentation masks.

B. subtilis:

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

StarDist model:

The StarDist 2D model was generated using the ZeroCostDL4Mic platform (Chamier et al., 2021). It was trained from scratch for 200 epochs (120 steps/epoch) on 155 paired image patches (image dimensions: (1024, 1024), patch size: (256,256)) with a batch size of 4, 10% validation data, 64 rays on grid 2, a learning rate of 0.0003 and a mae loss function, using the StarDist 2D ZeroCostDL4Mic notebook (v 1.12.2). Key python packages used include tensorflow (v 0.1.12), Keras (v 2.3.1), csbdeep (v 0.6.1), numpy (v 1.19.5), cuda (v 11.0.221). The training was accelerated using a Tesla P100GPU. The dataset was augmented by a factor of 3.

The model weights can be used in the ZeroCostDL4Mic StarDist 2D notebook, the StarDist Fiji plugin or the TrackMate Fiji plugin (v7+).

Author(s): Christoph Spahn^1,2, Mike Heilemann^1,3, Mia Conduit⁴, Séamus Holden^4,5, Pedro Matos Pereira^6,7, Mariana Pinho^6,8

Contact email: christoph.spahn@mpi-marburg.mpg.de, Seamus.Holden@newcastle.ac.uk, pmatos@itqb.unl.pt and mgpinho@itqb.unl.pt

Affiliation(s):

1) Institute of Physical and Theoretical Chemistry, Max-von-Laue Str. 7, Goethe-University Frankfurt, 60439 Frankfurt, Germany

2) ORCID: 0000-0001-9886-2263

3) ORCID: 0000-0002-9821-3578

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

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

6) Bacterial Cell Biology, Instituto de Tecnologia Química e Biológica António Xavier, Universidade Nova de Lisboa, Oeiras, Portugal

7) ORCID: 0000-0002-1426-9540

8) ORCID: 0000-0002-7132-8842

Sparsely Annotated Region and Organ Segmentation (SAROS) contributes a large heterogeneous semantic segmentation annotation dataset for existing CT imaging cases on TCIA. The goal of this dataset is to provide high-quality annotations for building body composition analysis tools (References: Koitka 2020 and Haubold 2023). Existing in-house segmentation models were employed to generate annotation candidates on randomly selected cases. All generated annotations were manually reviewed and corrected by medical residents and students on every fifth axial slice while other slices were set to an ignore label (numeric value 255). 900 CT series from 882 patients were randomly selected from the following TCIA collections (number of CTs per collection in parenthesis): ACRIN-FLT-Breast (32), ACRIN-HNSCC-FDG-PET/CT (48), ACRIN-NSCLC-FDG-PET (129), Anti-PD-1_Lung (12), Anti-PD-1_MELANOMA (2), C4KC-KiTS (175), COVID-19-NY-SBU (1), CPTAC-CM (1), CPTAC-LSCC (3), CPTAC-LUAD (1), CPTAC-PDA (8), CPTAC-UCEC (26), HNSCC (17), Head-Neck Cetuximab (12), LIDC-IDRI (133), Lung-PET-CT-Dx (17), NSCLC Radiogenomics (7), NSCLC-Radiomics (56), NSCLC-Radiomics-Genomics (20), Pancreas-CT (58), QIN-HEADNECK (94), Soft-tissue-Sarcoma (6), TCGA-HNSC (1), TCGA-LIHC (33), TCGA-LUAD (2), TCGA-LUSC (3), TCGA-STAD (2), TCGA-UCEC (1). A script to download and resample the images is provided in our GitHub repository: https://github.com/UMEssen/saros-dataset The annotations are provided in NIfTI format and were performed on 5mm slice thickness. The annotation files define foreground labels on the same axial slices and match pixel-perfect. In total, 13 semantic body regions and 6 body part labels were annotated with an index that corresponds to a numeric value in the segmentation file.

Body Regions

1. Subcutaneous Tissue
2. Muscle
3. Abdominal Cavity
4. Thoracic Cavity
5. Bones
6. Parotid Glands
7. Pericardium
8. Breast Implant
9. Mediastinum
10. Brain
11. Spinal Cord
12. Thyroid Glands
13. Submandibular Glands

Body Parts

1. Torso
2. Head
3. Right Leg
4. Left Leg
5. Right Arm
6. Left Arm

The labels which were modified or require further commentary are listed and explained below:

• Subcutaneous Adipose Tissue: The cutis was included into this label due to its limited differentiation in 5mm-CT.
• Muscle: All muscular tissue was segmented contiguously and not separated into single muscles. Thus, fascias and intermuscular fat were included into the label. Inter- and intramuscular fat is subtracted automatically in the process.
• Abdominal Cavity: This label includes the pelvis. The label does not separate between the positional relationships of the peritoneum.
• Mediastinum: The International Thymic Malignancy Group (ITMIG) scheme was used for the segmentation guidelines.
• Head + Neck: The neck is confined by the base of the trapezius muscle.
• Right + Left Leg: The legs are separated from the torso by the line between the two lowest points of the Rami ossa pubis.
• Right + Left Arm: The arms are separated from the torso by the diagonal between the most lateral point of the acromion and the tuberculum infraglenoidale.

For reproducibility on downstream tasks, five cross-validation folds and a test set were pre-defined and are described in the provided spreadsheet. Segmentation was conducted strictly in accordance with anatomical guidelines and only modified if required for the gain of segmentation efficiency.

The primary goal of the Semantic Drone Dataset is to enhance the safety of autonomous drone flight and landing procedures through improved semantic comprehension of urban environments. This dataset comprises imagery captured from a bird's-eye (nadir) perspective, showcasing over 20 houses, taken at altitudes ranging from 5 to 30 meters above the ground. The images are acquired using a high-resolution camera with a size of 6000x4000 pixels (24 megapixels). The training set includes 400 publicly accessible images, while the test set consists of 200 private images. Additionally, the dataset provides bounding box annotations for person detection within both the training and test sets.

The proposed dataset has been combined from three popular lung segmentation datasets: Darwin, Montgomery, and Shenzhen. The combined data allow researchers and clinicians to gain access to a good quality dataset, a large proportion of which has been manually annotated. The combined dataset consists of 6,810 images, with corresponding binary masks of lungs with the following distribution of images between the three datasets: • 6,106 images from the Darwin dataset; • 139 images from the Montgomery dataset; • 566 images from the Shenzhen dataset.

The Darwin dataset [1, 2] images include most of the heart, revealing lung opacities behind the heart, which may be relevant for assessing the severity of viral pneumonia. The lower-most part of the lungs, where visible, is defined by the extent of the diaphragm. Where present and not obstructive to the distinguishability of the lungs, the diaphragm is included up until the lower-most visible part of the lungs. A key property of this dataset is that image resolutions, sources, and orientations vary across the dataset, with the smallest image being 156x156 pixels and the largest being 5600x4700 pixels. Furthermore, we included the portable X-ray images which are of significantly lower quality as compared to standard X-rays. A key limitation of the Darwin dataset is that it does not contain lateral X-ray lung segmentations. It is worth noting that lung segmentations were performed by human annotators using Darwin's Auto-Annotate AI and then adjusted and reviewed by expert radiologists.

Both the Montgomery and Shenzhen datasets [3] were published by the United States National Library of Medicine and are made of posteroanterior chest X-ray images. These images are available to foster research in computer-aided diagnosis of pulmonary diseases with a special focus on pulmonary tuberculosis. The datasets were acquired from the Department of Health and Human Services (Maryland, USA) and Shenzhen №3 People's Hospital (Shenzhen, China). Both datasets contain normal and abnormal chest X-ray images with manifestations of tuberculosis and include associated radiologist readings.

References: 1. Darwin’s Auto-Annotate AI. Available: https://www.v7labs.com/automated-annotation 2. COVID-19 X-ray dataset. Available: https://github.com/v7labs/covid-19-xray-dataset 3. Jaeger S, Candemir S, Antani S, Wáng Y-XJ, Lu P-X, Thoma G. Two public chest X-ray datasets for computer-aided screening of pulmonary diseases. Quant Imaging Med Surg. 2014;4: 475–477. doi:10.3978/j.issn.2223-4292.2014.11.20

Needle Segmentation

## Overview

Needle Segmentation is a dataset for semantic segmentation tasks - it contains Gauge Needles annotations for 3,953 images.

## Getting Started

You can download this dataset for use within your own projects, or fork it into a workspace on Roboflow to create your own model.

  ## License

  This dataset is available under the [CC BY 4.0 license](https://creativecommons.org/licenses/CC BY 4.0).