WE3DS: RGB-D Stereo Images of Crops and Weeds with Hand-Annotated Semantic Segmentation Masks, 2,568 Images (BOKU Vienna, AgML Mirror)
- Provider
- Florian Kitzler, Norbert Barta, Reinhard W. Neugschwandtner, Andreas Gronauer and Viktoria Motsch (University of Natural Resources and Life Sciences, Vienna); Hugging Face mirror by Project-AgML
- License
- CC-BY-4.0 (source)
- Size
- 21 files
- Format
- Parquet (in sampled files)
- Stored at
- Hugging Face (https://huggingface.co/datasets/Project-AgML/WE3DS_weed_segmentation)
- Access
- Free and public. No account with the data provider is needed.
- DOI
- 10.5281/zenodo.7457983
- Category
- Agriculture & Environment
- Source
- https://huggingface.co/datasets/Project-AgML/WE3DS_weed_segmentation
About this dataset
WE3DS is an RGB-D image database for multi-class plant species semantic segmentation in crop farming: 2,568 images, each a color image plus a distance (depth) map, with hand-annotated ground-truth masks separating crop and weed species from soil. Images were taken under natural light with an RGB-D sensor made of two RGB cameras in a stereo setup, and the authors describe it as the first RGB-D dataset for multi-class plant species segmentation. The original train, validation and test split is kept in a split column. Useful for weed mapping, depth-aided segmentation, and selective weeding. Stored as Parquet shards with the image bytes and annotations in columns; load with datasets.load_dataset("Project-AgML/WE3DS_weed_segmentation") or through the AgML Python library (Project-AgML, UC Davis), which reformatted the original release. License: CC BY 4.0, as stated by the original publisher; cite Florian Kitzler, Norbert Barta, Reinhard W. Neugschwandtner, Andreas Gronauer and Viktoria Motsch et al. and the DOI 10.5281/zenodo.7457983.
How to load this dataset in Python
from datasets import load_dataset
ds = load_dataset("Project-AgML/WE3DS_weed_segmentation")
print(ds)Data structure
File types in sampled files: Parquet
Files
Example file: data/train-00000-of-00019.parquet
Analyze this dataset without downloading it
On Scigantic, WE3DS: RGB-D Stereo Images of Crops and Weeds with Hand-Annotated Semantic Segmentation Masks, 2,568 Images (BOKU Vienna, AgML Mirror) opens in a hosted JupyterLab notebook with the data already mounted, so you can read files where they are stored instead of copying them first. Each notebook includes an AI assistant that can see the mounted files and write and run analysis code. Open the dataset page to launch a notebook, or browse the full catalog.
This page is also available as Markdown and as JSON.
Related Agriculture & Environment datasets
- Soil E. coli Survival and Transfer to Cucumbers under Poultry Litter Amendments and Plastic Mulch: 2018 and 2019 Field Trials, Delaware (USDA ARS; Food AI Nexus Curation)
- Annotated Food Crop and Weed Images for Robotic Vision: 1,176 Images of 6 Crops and 8 Weed Species with 7,853 Boxes (AgML Mirror)
- OliveTreeCrownsDb: High-Resolution UAV Images of Olive Orchards with 66,950 Tree Crown Bounding Boxes (Sidi Mohamed Ben Abdellah University, AgML Mirror)
- Clay v1.5 NAIP-2
- AI4SmallFarms: Smallholder Crop Field Boundary Delineation from Sentinel-2, Vietnam and Cambodia (TACO)
- Litchi-UAV: Drone Images of Litchi Fruit in Dense Canopies with 39,415 Bounding Boxes, Guangzhou, China (AgML Mirror)
- Horseradish and Weed Detection Images from Commercial Fields in Southern Illinois: 2,697 Images with 25,298 YOLO Boxes (UIUC, AgML Mirror)
- VCD Vegetable Crops Dataset for Proximal Sensing: 2,801 Early-Stage Bean, Maize and Leek Images with 17,387 Boxes and Plant Structure Graphs (AgML Mirror)