DeepWeeds: A Multiclass Weed Species Image Dataset for Deep Learning

This repository makes available the source code and public dataset for the work, "DeepWeeds: A Multiclass Weed Species Image Dataset for Deep Learning", published with open access by Scientific Reports: https://www.nature.com/articles/s41598-018-38343-3. The DeepWeeds dataset consists of 17,509 images capturing eight different weed species native to Australia in situ with neighbouring flora. In our work, the dataset was classified to an average accuracy of 95.7% with the ResNet50 deep convolutional neural network.

The source code, images and annotations are licensed under CC BY 4.0 license. The contents of this repository are released under an Apache 2 license.

Download the dataset images and our trained models

images.zip (468 MB)

models.zip (477 MB)

Due to the size of the images and models they are hosted outside of the Github repository. The images and models must be downloaded into directories named "images" and "models", respectively, at the root of the repository. If you execute the python script (deepweeds.py), as instructed below, this step will be performed for you automatically.

TensorFlow Datasets

Alternatively, you can access the DeepWeeds dataset with TensorFlow Datasets, TensorFlow's official collection of ready-to-use datasets. DeepWeeds was officially added to the TensorFlow Datasets catalog in August 2019.

Weeds and locations

The selected weed species are local to pastoral grasslands across the state of Queensland. They include: "Chinee apple", "Snake weed", "Lantana", "Prickly acacia", "Siam weed", "Parthenium", "Rubber vine" and "Parkinsonia". The images were collected from weed infestations at the following sites across Queensland: "Black River", "Charters Towers", "Cluden", "Douglas", "Hervey Range", "Kelso", "McKinlay" and "Paluma". The table and figure below break down the dataset by weed, location and geographical distribution.

Data organization

Images are assigned unique filenames that include the date/time the image was photographed and an ID number for the instrument which produced the image. The format is like so: YYYYMMDD-HHMMSS-ID, where the ID is simply an integer from 0 to 3. The unique filenames are strings of 17 characters, such as 20170320-093423-1.

labels

The labels.csv file assigns species labels to each image. It is a comma separated text file in the format:

Filename,Label,Species ... 20170207-154924-0,jpg,7,Snake weed 20170610-123859-1.jpg,1,Lantana 20180119-105722-1.jpg,8,Negative ...

Note: The specific label subsets of training (60%), validation (20%) and testing (20%) for the five-fold cross validation used in the paper are also provided here as CSV files in the same format as "labels.csv".

models

We provide the most successful ResNet50 and InceptionV3 models saved in Keras' hdf5 model format. The ResNet50 model, which provided the best results, has also been converted to UFF format in order to construct a TensorRT inference engine.

resnet.hdf5 inception.hdf5 resnet.uff

deepweeds.py

This python script trains and evaluates Keras' base implementation of ResNet50 and InceptionV3 on the DeepWeeds dataset, pre-trained with ImageNet weights. The performance of the networks are cross validated for 5 folds. The final classification accuracy is taken to be the average across the five folds. Similarly, the final confusion matrix from the associated paper aggregates across the five independent folds. The script also provides the ability to measure the inference speeds within the TensorFlow environment.

The script can be executed to carry out these computations using the following commands.

To train and evaluate the ResNet50 model with five-fold cross validation, use python3 deepweeds.py cross_validate --model resnet.

To train and evaluate the InceptionV3 model with five-fold cross validation, use python3 deepweeds.py cross_validate --model inception.

To measure inference times for the ResNet50 model, use python3 deepweeds.py inference --model models/resnet.hdf5.

To measure inference times for the InceptionV3 model, use python3 deepweeds.py inference --model models/inception.hdf5.

Dependencies

The required Python packages to execute deepweeds.py are listed in requirements.txt.

tensorrt

This folder includes C++ source code for creating and executing a ResNet50 TensorRT inference engine on an NVIDIA Jetson TX2 platform. To build and run on your Jetson TX2, execute the following commands:

cd tensorrt/src make -j4 cd ../bin ./resnet_inference

Citations

If you use the DeepWeeds dataset in your work, please cite it as:

IEEE style citation: “A. Olsen, D. A. Konovalov, B. Philippa, P. Ridd, J. C. Wood, J. Johns, W. Banks, B. Girgenti, O. Kenny, J. Whinney, B. Calvert, M. Rahimi Azghadi, and R. D. White, “DeepWeeds: A Multiclass Weed Species Image Dataset for Deep Learning,” Scientific Reports, vol. 9, no. 2058, 2 2019. [Online]. Available: https://doi.org/10.1038/s41598-018-38343-3 ”

BibTeX

@article{DeepWeeds2019, author = {Alex Olsen and Dmitry A. Konovalov and Bronson Philippa and Peter Ridd and Jake C. Wood and Jamie Johns and Wesley Banks and Benjamin Girgenti and Owen Kenny and James Whinney and Brendan Calvert and Mostafa {Rahimi Azghadi} and Ronald D. White}, title = {{DeepWeeds: A Multiclass Weed Species Image Dataset for Deep Learning}}, journal = {Scientific Reports}, year = 2019, number = 2058, month = 2, volume = 9, issue = 1, day = 14, url = "https://doi.org/10.1038/s41598-018-38343-3", doi = "10.1038/s41598-018-38343-3" }

The DeepWeeds dataset consists of 17,509 images capturing eight different weed species native to Australia in situ with neighbouring flora.The selected weed species are local to pastoral grasslands across the state of Queensland.The images were collected from weed infestations at the following sites across Queensland: "Black River", "Charters Towers", "Cluden", "Douglas", "Hervey Range", "Kelso", "McKinlay" and "Paluma".

To use this dataset:

import tensorflow_datasets as tfds

ds = tfds.load('deep_weeds', split='train')
for ex in ds.take(4):
 print(ex)

See the guide for more informations on tensorflow_datasets.

https://storage.googleapis.com/tfds-data/visualization/fig/deep_weeds-3.0.0.png" alt="Visualization" width="500px">

Weeds are invasive plants that compete with crops for vital nutrients and often attract pests, significantly impacting agricultural productivity. They account for approximately 45% of the annual productivity loss in farming. Manual weeding methods, while effective, are labor-intensive and financially cumbersome, particularly for smallholder farmers. On the other hand, excessive reliance on chemical herbicides has led to herbicide resistance in several weed species, causing additional challenges in weed management.Emerging technologies, particularly artificial intelligence (AI) and computer vision, is revolutionizing the agricultural sector by automating labor-intensive tasks. Computer vision, in particular, uses advanced computational models to analyze visual cues from images, enabling the development of autonomous systems capable of performing tasks that often exceed the capabilities of the human visual system. In the context of robotic weeding, computer vision facilitates the precise identification of crops and weeds, allowing targeted herbicide application on weeds. To integrate these technologies high-quality datasets are critical component for development of accurate and robust models. So to address this need, the comprehensive MH-Weed16 Image Dataset is created from soybean fields of Maharashtra region located in India. Data acquisition was conducted from July 2023 to November 2023, ensuring a diverse natural field conditions. The dataset comprises total 18395 images of 16 weed species annotated under the guidance of subject-matter experts from agriculture universities. This dataset aims to serve as a foundational resource for training and evaluating machine learning and deep learning models. It will facilitate computer vision tasks of object detection and classification. Additionally, the dataset includes a total of 7,577 representative samples of crops along with weeds, categorized into three folders, with 6,656 weed samples annotated using bounding boxes. The images of crop with weed are captured from a top-down view to ensure accurate weed area estimation based on the bounding box areas. The MH-Weed16 dataset represents a significant step forward in the integration of technology for weed management strategies contributing towards sustainable agriculture practices

Weeds are undesired plants in agricultural fields that affect crop yield and quality by competing for nutrients, water, sunlight and space. Site-specific weed management (SSWM) through variable rate herbicide application and mechanical weed control have long been recommended in order to reduce the amount of herbicide and impact caused by uniform spraying. Accurate detection and classification of weeds in crop fields is a crucial first step for implementing such precise strategies. Drones are commonly used for image capturing but high wind pressure and different drone settings have a severe effect on the image quality, which potentially results in degraded images, e.g. due to motion blur. We publish a manually annotated and expert curated drone image dataset for weed detection in sorghum fields under challenging conditions. Our results show that our trained models generalize well regarding the detection of weeds, even for degraded captures due to motion blur. An UNet-like architecture with ResNet-34 as feature extractor achieved an F1-score of over 89 % on a hold-out test-set. Further analysis indicate that the trained model performed well in predicting the general plant shape, while most mis-classifications appeared at borders of the plants. Beyond that, our approach can detect intra-row weeds without additional information as well as partly occluded plants in contrast to existing research.

Github link: https://github.com/grimmlab/UAVWeedSegmentation

Please cite our original publication if you have used the data in your project or in any follow-up analysis (https://doi.org/10.1016/j.compag.2022.107388):

@article{GENZE2022107388, title = {Deep learning-based early weed segmentation using motion blurred UAV images of sorghum fields}, journal = {Computers and Electronics in Agriculture}, volume = {202}, pages = {107388}, year = {2022}, issn = {0168-1699}, doi = {https://doi.org/10.1016/j.compag.2022.107388}, url = {https://www.sciencedirect.com/science/article/pii/S0168169922006962}, author = {Nikita Genze and Raymond Ajekwe and Zeynep Güreli and Florian Haselbeck and Michael Grieb and Dominik G. Grimm}, keywords = {Deep learning, Weed detection, Weed segmentation, UAV, Precision agriculture}, }

The Weed Image Detection Dataset is a curated collection of images aimed at facilitating research and development in the field of automated weed detection and management. It comprises a diverse range of images featuring different types of weeds commonly found in agricultural and natural environments. These images are annotated to provide valuable information for training and evaluating computer vision models and algorithms.

Key Features:

Weed Varieties: The dataset encompasses a variety of weed species, including both common and invasive types, to ensure comprehensive coverage.

Image Diversity: Images are sourced from a wide range of locations, climates, and conditions to represent the real-world challenges of weed detection.

Annotations: Each image is meticulously annotated to indicate the presence and location of weeds, enabling the development and evaluation of object detection and segmentation models.

Image Quality: High-resolution images with varying lighting conditions and perspectives are included to simulate real-world scenarios.

Metadata: Additional metadata, such as location, date, and any available information about the plants, may be included in the dataset.

The Multiclass Weeds Dataset for Image Segmentation comprises two species of weeds: Soliva Sessilis (Field Burrweed) and Thlaspi Arvense L. (Field Pennycress). Weed images were acquired during the early growth stage under field conditions in a brinjal farm located in Gorakhpur, Uttar Pradesh, India. The dataset contains 7872 augmented images and corresponding masks. Images were captured using various smartphone cameras and stored in RGB color format in JPEG format. The captured images were labeled using the labelme tool to generate segmented masks. Subsequently, the dataset was augmented to generate the final dataset.

Crop And Weed Detection And Classification Using Computer Vision And Deep Learning Models

## Overview

Crop And Weed Detection And Classification Using Computer Vision And Deep Learning Models is a dataset for object detection tasks - it contains Crops And Weed annotations for 1,500 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).

Welcome to the Weed Detection Project, a groundbreaking initiative aimed at revolutionizing modern agriculture through the integration of advanced technologies. This project harnesses the power of deep learning to address the challenges posed by weed infestations in crop fields, offering a comprehensive solution for early detection and effective management.

Accurate and dependable weed detection technology is a prerequisite for weed control robots to do autonomous weeding. Due to the complexity of the farmland environment and the resemblance between crops and weeds, detecting weeds in the field under natural settings is a difficult task. Existing deep learning-based weed detection approaches often suffer from issues such as monotonous detection scene, lack of picture samples and location information for detected items, low detection accuracy, etc. as compared to conventional weed detection methods. To address these issues, WeedNet-R, a vision-based network for weed identification and localization in sugar beet fields, is proposed. WeedNet-R adds numerous context modules to RetinaNet’s neck in order to combine context information from many feature maps and so expand the effective receptive fields of the entire network. During model training, meantime, a learning rate adjustment method combining an untuned exponential warmup schedule and cosine annealing technique is implemented. As a result, the suggested method for weed detection is more accurate without requiring a considerable increase in model parameters. The WeedNet-R was trained and assessed using the OD-SugarBeets dataset, which is enhanced by manually adding the bounding box labels based on the publicly available agricultural dataset, i.e. SugarBeet2016. Compared to the original RetinaNet, the mAP of the proposed WeedNet-R increased in the weed detection job in sugar beet fields by 4.65% to 92.30%. WeedNet-R’s average precision for weed and sugar beet is 85.70% and 98.89%, respectively. WeedNet-R outperforms other sophisticated object detection algorithms in terms of detection accuracy while matching other single-stage detectors in terms of detection speed.

Here are a few use cases for this project:

1. Crop Management Systems: The weed detection model can be integrated into crop management software to help farmers quickly identify different types of weeds in their fields. This would allow them to apply the proper treatment in a timely manner, improving crop yield and reducing losses due to weed infestations.

2. Smart Agriculture Drones: The model could be used in smart drones for precision farming. These drones can scan fields, detect and identify the types of weed present, and selectively spray herbicides to eliminate those weeds without affecting the crop plants.

3. Botanical Research: Researchers studying plant species and biodiversity could use the model as a tool to identify and catalog various weed species in different environments. This could contribute to studies of invasive species and their impact on local ecosystems.

4. Gardening Apps: The model could be incorporated into gardening apps to assist users in identifying and treating weeds in their gardens. This would be particularly useful for gardeners without extensive knowledge of different weed species.

5. Environment Monitoring: Environmental agencies can use this model to monitor the presence and spread of invasive weed in natural parks and similar settings. Early detection and identification can aid in the proper management of these invasive species and protect the biodiversity of these areas.

Deep Weeds Segmentation Paper

## Overview

Deep Weeds Segmentation Paper is a dataset for instance segmentation tasks - it contains Deep Weeds annotations for 876 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 [MIT license](https://creativecommons.org/licenses/MIT).

The dataset CottonWeedDet12 consists of 5648 RGB images of 12-class weeds that are common in cotton fields in the southern U.S. states, with a total of 9370 bounding boxes. These images were acquired by either smartphones or hand-held digital cameras, under natural field light condition and throughout June to September of 2021. The images were manually labeled by qualified personnel for weed identification, and the labeling process was done using the VGG Image Annotator (version 2.10).

The dataset, at the time of publication, is the largest publicly available multi-class dataset dedicated to weed detection. It expects to facilitate communicate efforts to exploit state-of-the-art deep learning method to push weed recognition to the next level. With the WeedDet12 dataset, a performance benchmark of a suite of YOLO object detectors has been built for weed detection. Detailed documentation of the dataset, model benchmarking and performance results is given in an accompanying journal paper paper: Dang, F., Chen, D., Lu, Y., Li, Z., 2023. YOLOWeeds: A novel benchmark of YOLO object detectors for multi-class weed detection in cotton production systems. Computers and Electronics in Agriculture. https://doi.org/10.1016/j.compag.2023.107655

Every dataset in Weed-AI includes imagery of crops or pasture with weeds annotated, and is available in an MS-COCO derived format with standardised agricultural metadata.

Dataset

This dataset was created by Ryan Holbrook

Released under Data files © Original Authors

Contents

Deep Weeds Segm

## Overview

Deep Weeds Segm is a dataset for instance segmentation tasks - it contains Deep Weeds annotations for 886 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 [MIT license](https://creativecommons.org/licenses/MIT).

The dataset consists of orthophotos (build from UAV images) from a grassland field in which several Rumex obtusifolius plants were detected. The field is located in Germany (Kleve). The UAV images were acquired at 10, 15, and 30 meters height. Moreover, the labels/annotations from the Rumex obtusifolius plants in the images are also provided.

Context

The DeepWeedsX dataset consists of 17,508 unique 256x256 colour images in 9 classes. There are 15,007 training images and 2,501 test images. These images were collected in situ from eight rangeland environments across northern Australia.

Liaison with land care groups and property owners across northern Australia led to the selection of eight target weed species for the the collection of a large weed species image dataset; Chinee Apple (Ziziphus mauritiana), Lantana, Parkinsonia (Parkinsonia aculeata), Parthenium (Parthenium hysterophorus), Prickly Acacia (Vachellianilotica), Rubber vine (Cryptostegia grandiflora), Siam weed (Chromolaena odorata) and Snakeweed (Stachytarphetaspp).

DeepWeedsX is a subset of the DeepWeeds dataset, which was originally collected by Alex Olsen, and has previously been made openly accessible. We present a labeled variant with clearly defined training and test datasets. A validation dataset may be constructed for parameter optimization using a subset of the labeled training dataset.

Content

All class label files consist of Comma Seperated Values (CSVs) detailing the label and species, for example: 20161207-111327-0.jpg, 0 denotes that 20161207-111327-0.jpg belongs to class 0 (Chinee Apple).

Class and species labels are as follows:

0- Chinee Apple 1- Lantana 2- Parkinsonia 3- Parthenium 4- Prickly Acacia 5- Rubber Vine 6- Siam Weed 7- Snake Weed 8- Other.

All images are compressed in a single ZIP archive, and are labelled as per the class file labels.

Citation

To cite the DeepWeedsX dataset, kindly use the following BibTex entry:

@ARTICLE{8693488, author={C. {Lammie} and A. {Olsen} and T. {Carrick} and M. R. {Azghadi}}, journal={IEEE Access}, title={Low-Power and High-Speed Deep FPGA Inference Engines for Weed Classification at the Edge}, year={2019}, volume={}, number={}, pages={1-1}, keywords={Machine Learning (ML);Deep Neural Networks (DNNs);Convolutional Neural Networks (CNNs);Binarized Neural Networks (BNNs);Internet of Things (IoT);Field Programmable Gate Arrays (FPGAs);High-level Synthesis (HLS);Weed Classification}, doi={10.1109/ACCESS.2019.2911709}, ISSN={2169-3536}, month={},}

Acknowledgements

All original data collection was funded by the Australian Government Department of Agriculture and Water Resources Control Tools and Technologies for Established Pest Animals and Weeds Programme (Grant No. 4-53KULEI).

AI Weed Identification App Market Outlook

According to our latest research, the global AI Weed Identification App market size stood at USD 312 million in 2024 and is expected to reach USD 1.28 billion by 2033, growing at a robust CAGR of 17.2% during the forecast period. The marketÂ’s impressive growth trajectory is driven primarily by the increasing adoption of precision agriculture technologies and the urgent need for sustainable weed management solutions. The proliferation of smartphones, advancements in artificial intelligence, and heightened awareness among farmers regarding crop yield optimization are further catalyzing the expansion of this market. As per our latest research, the integration of AI with digital agriculture is transforming weed identification and control practices globally.

One of the principal growth factors for the AI Weed Identification App market is the escalating demand for improved crop productivity and cost-effective weed management. Traditional weed control methods, such as manual scouting and blanket herbicide application, are labor-intensive, time-consuming, and often result in excessive chemical usage. AI-powered weed identification apps offer a precise alternative by enabling real-time weed detection, species classification, and targeted herbicide application. This not only reduces operational costs for farmers but also minimizes environmental impact, aligning with the growing emphasis on sustainable agriculture. The rapid adoption of smartphones and high-speed internet connectivity in rural areas further facilitates the widespread use of these applications, making them accessible to a broader user base.

Another significant driver is the continuous advancement in artificial intelligence, particularly in computer vision and deep learning algorithms. These technological improvements have greatly enhanced the accuracy and reliability of weed identification apps, allowing them to distinguish between crops and a wide variety of weed species under diverse field conditions. AI-based apps can now process images captured via mobile devices or drones, providing instant analysis and actionable insights. This capability is particularly valuable in large-scale commercial farming, where timely and accurate weed detection can significantly impact overall yield and profitability. Moreover, the integration of these apps with other digital agriculture platforms, such as farm management systems and geospatial mapping tools, is creating a comprehensive ecosystem for precision weed management.

Government initiatives and regulatory support are also playing a pivotal role in propelling the AI Weed Identification App market forward. Many governments and agricultural organizations worldwide are promoting the adoption of digital technologies to enhance food security and environmental sustainability. Subsidies, training programs, and awareness campaigns are encouraging farmers and agronomists to embrace AI-driven weed management solutions. Additionally, the increasing focus on reducing chemical residues in food products and preserving soil health is prompting stakeholders across the agricultural value chain to invest in innovative weed control technologies. These collective efforts are fostering a conducive environment for the growth and adoption of AI weed identification apps.

From a regional perspective, North America currently leads the AI Weed Identification App market, accounting for the largest share in 2024. This dominance is attributed to the presence of advanced agricultural infrastructure, high technology adoption rates, and strong government support for precision farming initiatives. Europe follows closely, driven by stringent environmental regulations and a well-established research ecosystem. The Asia Pacific region is emerging as a high-growth market, fueled by rapid digitalization in agriculture, increasing smartphone penetration, and substantial investments in agri-tech startups. Latin America and the Middle East & Africa are also witnessing steady growth, albeit at a slower pace, as awareness and infrastructural capabilities continue to improve.

The introduction of the AI Weed Identification Robot is revolutionizing the way farmers approach weed management. Unlike traditional methods, this robotic solution leverages advanced AI algorithms to autonomously navigate fields, i

The authentic labels, predicted results and confusion matrices of all models (AlexNet, GoogLeNet, VGG16, SqueezeNet, Xception, and NasNet-Mobile)

This dataset includes data on the chemical, physical and biological traits of weed seeds of 11 arable weed species in relation to the persistence of these seeds in the soil seedbank within a common-garden burial study. We performed a common garden weed seed burial study at the University of Illinois Crop Sciences Research and Education Center in Savoy, IL (40.048757 N, -88.237206 E), from October 2007 through October 2012. The experiment was arranged in a split-plot design with four replications of the sub-plot variable species nested within main plot variable burial duration (1 to 5 years). Eleven annual weed species were included, spanning a broad range of seed sizes, dormancy types and seedbank persistence: Abutilon theophrasti Medik (velvetleaf), Ambrosia trifida L. (giant ragweed), Amaranthus tuberculatus [Moq]. Sauer (common waterhemp), Bassia scoparia [L.] A. J. Scott (kochia), Chenopodium album L. (common lambsquarters), Ipomoea hederacea Jacq. (ivyleaf morningglory), Panicum miliaceum L. (wild proso millet), Polygonum pensylvanicum L. (Pennsylvania smartweed), Setaria faberi Herrm. (giant foxtail), Setaria pumila [Poir] Roem. (yellow foxtail) and Thlaspi arvense L. (field pennycress). Weed seeds were collected in 2007 from the experimental site and adjoining fields by gently shaking mature inflorescences over a bucket and bulking seeds from multiple plants to form a composite sample for each species. Light seed were removed by processing with a seed cleaner, after which seeds were stored in air tight containers at 4C until burial. Immediately prior to burial, seed viability was assayed with tetrazolium. Burial units consisted of 100 seeds of a given species placed in the bottom of a 2.5 cm deep square tray, 10 cm on a side, made of 0.5 mm stainless steel wire mesh. Tray bottoms were permeable to water, but prevented seeds from escaping. Trays were filled 2 cm deep with soil from a nearby grass sward that had not been cropped for over 30 years, to avoid contamination with weed seeds (verified by elutriating samples of this soil). Within each experimental unit, we excavated a 2 cm deep rectangle 30 cm wide by 40 cm long, and placed trays for each of the 11 species side by side into this depression so that their soil surface was flush with the surrounding soil, leaving a 0.5 cm wire mesh lip exposed in each tray. Each experimental unit was covered by wire mesh with 1 cm square openings to permit access to invertebrate granivores. The study plot was fenced to exclude large vertebrates. Seedling emergence was recorded weekly from March through October every year. Seed trays for a given burial duration treatment were removed in October of the assigned year and seeds recovered via elutriation (Wiles et al. 1996). Recovered seeds were incubated under oscillating temperature conditions (15 C/dark for 10 hr, 25 C/light for 14 hr) for 2 weeks and germination recorded. Ungerminated seeds assessed as viable through tetrazolium testing were considered dormant. SEED TRAITS We measured chemical and physical seed traits on freshly collected seeds following the methods outlined in Tiansawat et al. (2014), using multiple measures of each trait class to provide functional redundancy and allow them to be treated as latent or manifest variables during multivariate analyses. For the chemical defense trait class we measured ortho-dihydroxyphenol (o-DHP) concentration, abundance and diversity of phenolic compounds quantified with high performance liquid chromatography, impact of seed homogenate on brine shrimp survival, and seed removal by invertebrate granivores. Physical traits measured included seed coat thickness, seed mass, and seed coat rupture force. Pairwise interspecific phylogenetic distances were quantified using the phydist subroutine of Phylocom 4.2 (www.phylodiversity.net). Also included is a list of references from the associated literature review. Resources in this dataset:Resource Title: Weed seed defense traits. File Name: Davis et al 2016_Seed Persistence.xlsxResource Description: This data set contains information on weed seed chemical, biological and physical traits in relation to weed seed persistence in the soil seedbank, as measured through a common garden burial study in Urbana, IL, from 2007 through 2012.Resource Title: Data Dictionary. File Name: DataDictionary.csvResource Description: Describes variables and units for each worksheet: Seed Persistence; Mean Seed Persistence vs. Traits; Literature Review of Dormancy vs. Persistance.

Type of data: RGB digital images Data format: .jpg Number of images: 3632 Number of classes: 11 How data were acquired : Through a smartphone camera Data source location: Rice fields of Narsingdi and Munshiganj, Bangladesh Where applicable: Deep learning, image processing, agricultural studies