CWD30: A Comprehensive and Holistic Dataset for Crop Weed Recognition in Precision Agriculture

Abstract

The growing demand for precision agriculture necessitates efficient and accurate crop-weed recognition and classification systems. Current datasets often lack the sample size, diversity, and hierarchical structure needed to develop robust deep learning models for discriminating crops and weeds in agricultural fields. Moreover, the similar external structure and phenomics of crops and weeds complicate recognition tasks. To address these issues, we present the CWD30 dataset, a large-scale, diverse, holistic, and hierarchical dataset tailored for crop-weed recognition tasks in precision agriculture. CWD30 comprises over 219,770 high-resolution images of 20 weed species and 10 crop species, encompassing various growth stages, multiple viewing angles, and environmental conditions. The images were collected from diverse agricultural fields across different geographic locations and seasons, ensuring a representative dataset. The dataset's hierarchical taxonomy enables fine-grained classification and facilitates the development of more accurate, robust, and generalizable deep learning models. We conduct extensive baseline experiments to validate the efficacy of the CWD30 dataset. Our experiments reveal that the dataset poses significant challenges due to intra-class variations, inter-class similarities, and data imbalance. Additionally, we demonstrate that minor training modifications like using CWD30 pretrained backbones can significantly enhance model performance and reduce convergence time, saving training resources on several downstream tasks. These challenges provide valuable insights and opportunities for future research in crop-weed recognition. We believe that the CWD30 dataset will serve as a benchmark for evaluating crop-weed recognition algorithms, promoting advancements in precision agriculture, and fostering collaboration among researchers in the field.

Full text

1
CWD30: A Comprehensive and Holistic Dataset for
Crop Weed Recognition in Precision Agriculture
Talha Ilyas∗† , Dewa Made Sri Arsa∗‡, Khubaib Ahmad∗ †, Yong Chae Jeong∗, Okjae Won§, Jong Hoon Lee†, and
Hyongsuk Kim†
∗Division of Electronics and Information Engineering, Jeonbuk National University, Jeonju 54896, Republic of
Korea
†Core Research Institute of Intelligent Robots, Jeonbuk National University, Jeonju 54896, Republic of Korea
‡Department of Information Engineering, Universitas Udayana, Bali, Indonesia
§Production Technology Research Division, Rural Development Administration, National Institute of Crop
Science, Miryang, Republic of Korea
Abstract—The growing demand for precision agriculture ne-
cessitates efficient and accurate crop-weed recognition and clas-
sification systems. Current datasets often lack the sample size,
diversity, and hierarchical structure needed to develop robust
deep learning models for discriminating crops and weeds in
agricultural fields. Moreover, the similar external structure and
phenomics of crops and weeds complicate recognition tasks. To
address these issues, we present the CWD30 dataset, a large-scale,
diverse, holistic, and hierarchical dataset tailored for crop-weed
recognition tasks in precision agriculture. CWD30 comprises over
219,770 high-resolution images of 20 weed species and 10 crop
species, encompassing various growth stages, multiple viewing
angles, and environmental conditions. The images were collected
from diverse agricultural fields across different geographic loca-
tions and seasons, ensuring a representative dataset. The dataset’s
hierarchical taxonomy enables fine-grained classification and
facilitates the development of more accurate, robust, and gen-
eralizable deep learning models. We conduct extensive baseline
experiments to validate the efficacy of the CWD30 dataset. Our
experiments reveal that the dataset poses significant challenges
due to intra-class variations, inter-class similarities, and data
imbalance. Additionally, we demonstrate that minor training
modifications like using CWD30 pretrained backbones can sig-
nificantly enhance model performance and reduce convergence
time, saving training resources on several downstream tasks.
These challenges provide valuable insights and opportunities for
future research in crop-weed detection, fine-grained classification,
and imbalanced learning. We believe that the CWD30 dataset
will serve as a benchmark for evaluating crop-weed recognition
algorithms, promoting advancements in precision agriculture,
and fostering collaboration among researchers in the field. The
data is available at: https://github.com/Mr-TalhaIlyas/CWD30
Index Terms—precision agriculture, crop weed recognition,
benchmark dataset, plant growth stages, deep learning.
I. INTROD UCTION
PRE CISION agriculture is essential to address the increasing
global population and the corresponding demand for a
70% increase in agricultural production by 2050 [1]. The
challenge lies in managing limited cultivation land, water
scarcity, and the effects of climate change on productivity.
One critical aspect of precision agriculture is the effective
control of weeds that negatively impact crop growth and yields
by competing for resources and interfering with crop growth
through the release of chemicals [2]–[4].
Recent advances in deep learning have revolutionized the
field of computer vision, with Convolutional Neural Networks
(CNNs) and transformers becoming the backbone of numerous
state-of-the-art models [5]–[7]. However, their performance
relies heavily on the quality and diversity of training data [8],
[9], emphasizing the importance of comprehensive agricultural
datasets for model development [10]. But the agricultural
domain often suffers from the deficiency of task-specific
data [10], [11]. Which can result in insufficient data variety,
overfitting, inadequate representation of real-world challenges,
and reduced model robustness. These limitations hinder the
model’s ability to generalize and accurately recognize crops
and weeds in diverse real-world situations. To overcome these
issues, researchers employ techniques like data augmentation
[12], [13], transfer learning [14] , or synthetic data generation
[15], although these approaches may not always achieve the
same performance level as models trained on larger, more
diverse datasets [16]. Transfer learning (fine-tuning) [17] is a
common approach for training deep learning models in agri-
culture, as it involves using pretrained weights from other tasks
(e.g., ImageNet) to address data deficiency [18]. Pretrained
weights from ImageNet [19] and COCO [20] are commonly
used but are less suitable for domain-specific agricultural
tasks due to their generic content [10], [21]. Thus absence
of a centralized benchmark repository for agriculture-specific
datasets hinders the development of computer-aided precision
agriculture (CAPA) systems.
In this study, we introduce and evaluate the crop weed
recognition dataset (CWD30) dataset, a large-scale and diverse
collection of various crops and weed images that captures the
complexities and challenges of real-world precision agriculture
scenarios. The CWD30 dataset comprises a collection of
219,770 images that encompass 10 crop classes and 20 weed
classes. These images capture various growth stages, multiple
viewing angles, and diverse environmental conditions. Figure
1 shows some image samples, while Figure 2 displays the
number of images per category. The CWD30 dataset addresses
the significant intra-class difference and large inter-species
similarity of multiple crop and weed plants. We train various
deep learning models, including CNNs and transformer-based
architectures, on the CWD30 dataset to assess their perfor-

2
Fig. 1. Crop and Weed image samples from CWD30 dataset, captured at different life cycle stages, under varying environment and from different viewing
angles.Key elements in the images are highlighted: pink-bordered images represent similarities at a macro class level (crop vs weed); orange boxes indicate
the variability within a single weed species due to environmental factors such as indoor vs outdoor settings and soil type; images encased in red and brown
borders demonstrate visually similar crop and weed classes; images marked with black dashed lines represent weeds cultivated in a laboratory setting; small
inset boxes on each image provide information about the weather conditions and camera angle and plant age at time of capture.
mance and investigate the impact of pretraining. Furthermore,
we analyze the structure of the feature embeddings obtained by
these models and compare their performance on downstream
tasks, such as pixel-level crop weed recognition
In summary, building upon the aforementioned challenges
and limitations we make the following main contributions:
•We present the crop-weed dataset (CWD30), which, to
the best of our knowledge, is the first truly holistic, large-
scale crop weed recognition dataset available to date.
•Proposed dataset encompasses a wide range of plant
growth stages, i.e., from seedlings to fully mature plants.
This extensive coverage of growth stages ensures that
the CWD30 dataset captures the various morphologi-
cal changes and developmental stages plants undergo
throughout their life cycle. By incorporating these diverse
growth stages, the dataset provides a more comprehensive
representation of real-world agricultural scenarios. Con-
sequently, deep learning models trained on this dataset
can better adapt to the inherent variability in plant appear-
ances and growth stages, Figure 7a shows a few samples
of plants at various growth stages.
•The CWD30 dataset offers a unique advantage by includ-
ing multi-view images, captured at various angles. This
comprehensive representation of plants account for vari-
ous viewpoints and lighting conditions, which enhances
the dataset’s ability to model real-world situations. The
multi-view images enable the development of more robust
and generalizable deep learning models, as they allow the

3
models to learn from a broader range of visual features
and better understand the complexities and variations
commonly found in real-field settings (see section III for
details).
•Compared to existing agricultural datasets that focus on
specific plant parts like branches or leaves, the proposed
CWD30 dataset offers high-resolution images of entire
plants in various growth stages and viewpoints. This
comprehensive nature of the CWD30 dataset allows for
the generation of simpler, plant-part-specific datasets by
cropping its high-resolution images. As a result, the
CWD30 dataset can be considered a more versatile
and complete resource compared to existing datasets.
This dataset contributes to overcoming the limitations
of previous datasets and advances the field of precision
agriculture.
•Additionally, we demonstrate that models pretrained
on the CWD30 dataset consistently outperform their
ImageNet-1K pretrained counterparts, yielding more
meaningful and robust feature representations. This im-
provement, in turn, enhances the performance of state-of-
the-art models on popular downstream agricultural tasks
(see section V for details).
These contributions can further advance the research and
development of reliable CAPA systems.
The rest of this article unfolds as follows: Section II
provides a review of related literature and relevant datasets.
Section III explains the development of the CWD30 dataset,
its unique characteristics, and draws comparisons with other
agricultural datasets. The experimental setup is outlined in
Section IV. Following this, Section V delves into the analysis
of experimental results and the inherent advantages offered by
the CWD30 dataset. Finally, we wrap up the article in the
conclusion.
II. RE LATE D WORKS
A. Crop Weed Recognition
Crop-weed recognition is vital in CAPA systems for effi-
cient and sustainable farming practices. Reliable recognition
and differentiation allow for effective weed management and
optimal crop growth, reducing chemical usage and minimizing
environmental impact [8], [22]. It also helps farmers oversee
their crops’ health, enabling prompt response and lowering
the possibility of crop loss from weed infestations [5], [23].
However, these systems face limitations due to the reliance
on small datasets [24], resulting in reduced model robustness,
overfitting, and inadequate representation of real-world chal-
lenges.
Several studies have shown the potential of deep learning
techniques in addressing key components and challenges in
developing CAPA systems, such as unmanned weed detec-
tion [39], fertilization [40], irrigation, and phenotyping [41].
Kamilaris et al. [22] conducted experiments that showed deep
learning outperforming traditional methods. Westwood et al.
[42] discussed the potential of deep learning-based plant
classification for unmanned weed management. Wang et al.
highlighted the main challenges in differentiating weed and
TABLE I
COMPARATIVE AN ALYSI S OF VARI OUS AG RI CULTU RA L DATAS ETS : KE Y ATTRI BUT ES AN D CHA RACT ERI ST ICS .TH E SYM BO L ‘˜’ IND ICAT ES AN A PPRO XIM ATE VALUE . HH, DM, AN D VM CO RRE SPO ND TO
HANDHELD,DEVICE MOUNTED,AND V EH ICL E MOU NT ED CA MER AS,R ES PEC TIV ELY.
Dataset #Images # Cat. Coverage Environment Background Avg. Image
Resolution
Multi
View
Growth
Stages Availability Image Content Location Acquisition
Platform
Deep Weeds [25] 17,509 9 weeds outdoor complex 256x256
No No
public Full plant Roadside Tripod /
Overhead
Camera
Plant Seedling [26] 5,539 12 weeds indoor simple ∼355x355 public Full plant Trays
Fruit Leaf [27] 4,503 12 fruits indoor simple 6000x4000 public single leaf tray
PDDB [28] 46,409 56 Fruits, crops indoor simple 2048x1368
No No
public
Single leaf Lab
Handheld
RGB
Camera
Corn2022 [29] 7,701 4 corn outdoor Simple 224x224 public
LWDCD2020 [30] 12,160 10 wheat outdoor simple 224x224 private
Plant Village [31] 54,309 38 fruits, crops indoor simple 256x256 No No public Single leaf Lab
Plant Doc [32] 2,598 17 fruits, crops outdoor complex ∼1070x907
RiceLeaf [33] 5,932 4 rice outdoor simple 300x300
No No
private Single leaf
Farmland
CLD [34] 15,000 6 cassava outdoor complex 800x600 public Single branch
AppleLeaf [35] 23,249 6 frutis outdoor simple 4000x2672 public Single leaf
CNU [36] 208,477 21 weeds outdoor complex - private Single branch
PDD271 [37] 220,592 271 fruits, crops,
vegetables outdoor Simple 256x256 private Single leaf
IP102 [38] 75,222 102 crop pests Simple /
complex Simple ∼525x413 No No private Single pest
on leaf
Farmland, sketch,
drawings
Search
Engines
CWD30 219,778 30 crops, weeds Simple /
complex
Simple /
complex
∼4032x3024 Yes Yes public Full plant Farmland, Pots
HH /DM /
VM / Overhead
camera

4
Fig. 2. A comparative plot of class distributions per viewing angle. Numbers
in parenthesis represent the total number of images of that plant category.
crop species in CAPA systems. Moreover, Wang et al. [43]
and Khan et al. [44] emphasized the importance of combining
spectral and spatial characteristics for remote sensing and
ground-based weed identification approaches. Hasan et al. [8]
conducted a comprehensive survey of deep learning techniques
for weed detection and presented a taxonomy of deep learning
techniques.
However, recent studies by Moazzam et al. [4] and Coleman
et al. [45] identified research gaps, such as a lack of substantial
crop-weed datasets and generalized models and concluded that
methods like data augmentation and transfer learning might
not always produce results on par with models trained on more
substantial, diverse datasets. To address these limitations and
challenges, further research is needed to improve the accuracy
and robustness of CAPA systems. Considering the identified
research gaps and challenges, this work presents the CWD30
dataset, specifically designed to address the limitations of
existing agricultural datasets. Our aim is to facilitate the de-
velopment of accurate and reliable CAPA systems, ultimately
Fig. 3. Visual comparison of CWD30 dataset with other related datasets.
enhancing the effectiveness and sustainability of precision
agriculture practices.
B. Related Datasets
Here we provide an overview of several related agricul-
tural datasets that have been previously proposed for crop-
weed recognition and other agricultural tasks [25]–[38]. These
datasets, while valuable, have certain limitations that the
CWD30 dataset aims to address.
Plant Seedling:The Plant Seedlings Dataset [26] features
approximately 960 unique plants from 11 species at various
growth stages. It consists of annotated RGB images with a
resolution of around 10 pixels per mm. Three public versions
of the dataset are available: original, cropped, and segmented.
For comparison in this study, we use the cropped plants v2
version, which contains 5,539 images of 12 different species.
The dataset is imbalanced, with some classes having up to 654
samples (chickweed) and others as few as 221 (wheat).
The dataset was collected over 20 days (roughly 3 weeks)
at 2-to-3-day intervals in an indoor setting. Plants were grown
in a styrofoam box, and images were captured using a fixed
overhead camera setup. This database was recorded at the
Aarhus University Flakkebjerg Research station as part of a
collaboration between the University of Southern Denmark
and Aarhus University.
CNU: This weeds dataset from Chonnam National Univer-
sity (CNU) in the Republic of Korea [36] consists of 208,477
images featuring 21 species. Captured on farms and fields

5
Fig. 4. Taxonomy of CWD30 dataset. Showcasing the hierarchical organization of crop and weed species included in the dataset.
using high-definition cameras, the images encompass various
parts of weeds, including flowers, leaves and fruits. A visual
comparison between the CNU dataset and the CWD30 dataset
is illustrated in a Figure 3. However, unlike the CWD30
dataset, the CNU dataset does not encompass the growth stages
and multiple viewing angles. The CNU dataset is imbalanced,
with over 24,300 images of shaggy soldier and only about 800
images of spanish needles.
Deep Weeds: The Deep Weeds [25] dataset consists of
17,509 low-resolution images of herbaceous rangeland weeds
from 9 species. This dataset features a minimum of 1009
images and a maximum of 9016 images per category.
IP102: Wu et al. [38] developed the IP102 dataset to
further insect pest recognition research in computer vision.
They initially gathered over 300,000 images from popular
search engines, which were then labeled by volunteers to

6
TABLE II
LIS T OF WE ED SP ECI ES I NCL UDE D IN TH E CWD30 DATASE T,THE IR GE OGR AP HIC AL DI STR IBU TI ON,A ND TH E CRO P SPE CIE S TH EY CO MMO NLY AFF ECT,
EM PHA SIZ ING T HE I MPO RTANC E OF WE ED RE CO GNI TIO N AND M ANAG EM ENT I N GLO BAL AG RI CULTU RE [46], [47].
Weed Name Countries Found In Crops Affected
Cockspur Grass United States, Canada, Europe, Asia, Australia, Africa Corn, millets
Early Barnyard Grass Europe, Asia, Africa Corn, millets
Fall panicum North America, Europe, Asia Corn, millets
Fingergrass Worldwide Corn, millets
Green foxtail North America, Europe, Asia Corn, millets
Indian goosegrass Asia, Africa, South America Corn, millets
Poa annua Worldwide Corn, millets
Copper leaf Worldwide Corn, millets, beans, peanuts
Goosefoot Worldwide Corn, millets, beans
Henbit North America, Europe, Asia Corn, millets, beans
Livid pigweed North America, Europe, Asia Corn, millets, beans
Purslane Worldwide Corn, millets, beans
Redroot pigweed North America, Europe, Asia Corn, millets, beans
Smooth pigweed North America, Europe, Asia Corn, millets, beans
White goosefoot North America, Europe, Asia Corn, millets, beans
Asian flats edge Asia, North America, South America Millets, beans
Bloodscale sedge North America, Europe, Asia Millets
Nipponicus sedge Asia, Europe, North America Millets
Korean dock Asia, North America, Europe Millets, beans, sesame
Asiatic dayflower Asia, North America, Europe, Australia Millets, beans, sesame
TABLE III
GLO BAL PR ODU CTI ON SH AR E ,IN MI LLI ON M ETR IC TO NS (M), OF T HE 10 C ROP S PEC IE S INC LUD ED IN T HE CWD30 DATAS ET FO R THE Y EAR 2020 T O
2021, ACRO SS VARI OU S COU NTR IES ,EM PHA SIZ IN G THE IR SI GNI FIC ANC E AND C ONT RI BUT ION T O WOR LDW IDE AG RI CULTU RA L PROD UC TIO N [48]–[50].
Country Corn Foxtail Millet Great Millet Proso Millet Bean Green Gram Peanut Red Bean Sesame
United States 358.4M - 9.7M - - - 2.79M - -
China 260.8M 6.5M - 1.8M - 0.6M 17.9M 2.2M -
Brazil 81M - - - 4.2M - - - -
India 31.65 1M 6M 2.2M 6.5M 2M 6.7M - 0.8M
Nigeria 12.4 - 7.1M - - - 4.23M - -
Myanmar - - - - 3.9M 0.9M - - 0.6M
Russia 13.87 - - 1.1M - - - - -
Japan - - - - - - - 0.2M -
South Korea - - - - - - - 0.1M -
Sudan - - - - - - - - 1.1M
Share of Global
Production (%)
67.1 83.3 39.7 73.6 48.7 87.5 62.9 65.7 58.3
ensure relevance to insect pests. Following a data cleaning
process, the IP102 dataset consisted of about 75,000 images
representing 102 species of common crop insect pests. The
dataset also captures various growth stages of some insect pest
species.
PDD271: Liu et al. [37] developed a large-scale dataset
to support plant disease recognition research, consisting of
220,592 images across 271 disease categories. The data was
collected in real farm fields with a camera-to-plant distance
of 20-30 cm to ensure consistent visual scope. The dataset
consists of a minimum of 400 images per category and a
maximum of 2000 images.
Researchers are actively working on plant recognition,
frequently utilizing image databases containing samples of
particular species to evaluate their methods. The creation of a
database necessitates significant time, planning, and manual la-
bor [26], [51]. Data is usually captured using an array of equip-
Fig. 5. Schematic representation of the file naming convention in the CWD30
dataset, with each segment separated by “ ” indicating specific information
about the image, such as species, growth stage, camera angle, and unique ID.
ment, from readily available commercial cameras to custom-
built sensors designed for specific data acquisition tasks [52],
[53]. Consequently, data collected by various researchers differ
in quality, sensor type, and quantity, as well as encompassed

7
Fig. 6. Illustration of camera placement for capturing images at various
angles, along with sample images captured at those angles under different
weather conditions.
distinct species. This leads to a diverse and occasionally sparse
dataset, often tailored for highly specialized research.
Compared to previous datasets, our proposed CWD30
dataset is unique in that it not only includes images captured
from multiple angles, at various growth stages of the plant
under varying weather conditions, but also features full plant
images rather than just parts of plants (like leaves or branches)
see Figure 3. This allows deep learning models to learn more
robust and holistic features for better recognition, differen-
tiation, and feature extraction. By addressing the domain-
specific challenges of real-field agricultural environments and
providing a diverse, varied, and extensive collection of images,
CWD30 not only advances research in the field, but also
enhances data efficiency and performance in a wide range of
downstream agricultural tasks.Table I presents the statistical
information for various agriculture-related datasets.
III. DATA COL LECTI ON , PR EP ROC ES SING AND
PROP ERT IE S
In this section, we provide a detailed explanation of the col-
lection process, preprocessing, and properties of the proposed
CWD30 dataset.
A. Taxonomic System Establishment
We developed a hierarchical taxonomic system for the
CWD30 dataset in collaboration with several agricultural ex-
perts from the Rural Development Authority (RDA) in the
Republic of Korea. We discussed the most common weed
species that affect economically significant crops globally [46],
[47]. A summary of these weeds, the crops they impact, and
the countries where they are prevalent is provided in Table II.
We ultimately chose to collect data on approximately 20 of
the most problematic weed species worldwide. The selection
of the 10 crops included in the CWD30 dataset was based on
their share in global production and regional importance [48]–
[50], ensuring the dataset’s relevance and applicability in real-
world precision agriculture scenarios. Table III indicates that
these crops have considerable shares of global production, with
percentages ranging from 39.7% to 87.5%. By incorporating
crops with substantial importance across various countries, the
CWD30 dataset establishes a taxonomy system that addresses
the needs of diverse agricultural environments and promotes
research in crop recognition and management.
For weed species not native to Korea, the RDA cultivated
them in pots within their facility, as shown in Figure 1 (dashed
black borders). As for the selected crops, they were divided
into two subcategories based on their primary commercial
value: economic crops (EC) and field crops (FC). Field crops
include staples such as corn and millet, while economic crops
encompass legumes (e.g., beans) and oilseeds (e.g., sesame).
The resulting hierarchical structure is illustrated in Figure 4.
Each crop is assigned both a micro and macro-class based
on its properties, whereas weeds are only assigned a micro-
class, such as grasses, broad leaves, or sedges. In the CWD30
dataset, we also include a hold-out test set consisting of 23,502
mixed crop and weed (MCW) images, captured both indoors
and outdoors, to facilitate the validation of developed models,
see Figure 2. We have included a comprehensive table in the
appendix of this paper, providing a detailed taxonomy for each
plant species within the CWD30 dataset which explains the
hierarchical classification, right from the domain, kingdom,
and phylum, down to the order, family, genus, and species of
each plant.
B. Data Collection
To assemble a benchmark database, we formed five teams:
four dedicated to collecting images in farms and fields, and
one focused on gathering images from RDA’s research facility.
Each team was composed of three students from our institute
and one field expert. The image collection devices provided
to each team varied, including three Canon-SX740 HS, three
Canon EOS-200D, three Android phone-based cameras, three
iPhone-based cameras, and one DJI Mavic Pro 2.
Each team was tasked with capturing images of two crops
and four weeds twice a week. The full dataset is collected
over a span of three years from 2020 to 2022. Since image
collection is a manual process, the data recorded by different
team members varied in quality, perspective, height, sensor
type, and species. To ensure diverse data collection, we shuf-
fled the teams monthly and assigned them to collect images
of different crops and weeds. This approach helped us obtain
a diverse dataset that covers a wide spectrum of real-world
challenges and domain shifts, stemming from different sensor
types, field environments, and fields of view. Figure 1 shows
samples of the collected images.
C. Data Filtering, Labelling and Distribution
The entire data construction process spanned three years.
Alongside image collection, five experts reviewed each image
to ensure label accuracy monthly. They then removed blurry
and noisy images to maintain a clean dataset. The resulting
CWD30 dataset comprises 219,778 images, 10 crop types, and
20 weed species. The distribution of each species is depicted in
Figure 2. The minimum number of images per species is 210,
while the maximum is 12,782. This unbalanced distribution
reflects real-world scenarios where it is challenging to obtain
data samples for certain classes. In our case, this occurred for

8
(a) (b)
Fig. 7. (a) A visual representation of plant growth stages, spanning an 8-week period from seedling to maturity, showcasing the developmental progression,
changes in color, shape and texture of the plant over time. (b)Radar graph illustrating the distribution of images in the CWD30 dataset across each growing
stage during the 8-week period.
weed species that were difficult to cultivate in Korea’s weather
conditions. As for labeling, each file is saved with a unique
naming format, an example of which can be seen in Figure 5.
D. Data Splits
The CWD30 dataset comprises 219,778 images and 30
plant species. To ensure more reliable test results, we em-
ployed a K-fold validation method with K=3, guaranteeing
enough samples for each category in the testing set [54].
We divided the data into three randomized folds for training
(74,724), validation (72,526), and testing (72,526), adhering
to a 0.33:0.33:0.34 split ratio. For each fold, we partitioned
every plant species into three sections, taking care to include
an equal proportion of the smallest class within each section
(refer to Figure 2). The training, validation, and testing sets
were split at the micro-class level.
E. Viewing Angles and Growth Stages
Our proposed CWD30 dataset stands out from previous
datasets due to its unique and beneficial properties, with three
Fig. 8. Data imbalance ratio (IR) of proposed dataset in comparison with
other datasets.
prominent features: (i) images captured from multiple angles,
(ii) images taken at various growth stages and under varying
weather conditions, and (iii) full plant images instead of just
plant parts like leaves or branches. These characteristics enable
deep learning models to learn more robust and comprehensive
features for enhanced recognition, differentiation, and feature
extraction.
Capturing plant images from different angles for deep
learning models results in robust feature learning, improved
occlusion handling, scale and rotation invariance, and better
management of lighting and shadow variations. This leads to
more accurate and reliable CAPA systems that perform well in
real-world agricultural environments. Figure 6 depicts a visual
representation of the various angles used for image collection.
Furthermore, the growing interest in plant phenomics and
the use of image-based digital phenotyping systems to measure
morphological traits has led to increased efforts to bridge the
genotyping-phenotyping gap. However, research in this area is
limited, mainly due to the lack of available datasets providing
whole-plant level information rather than specific plant parts,
such as leaves, nodes, stems, and branches. The CWD30
dataset, which includes full plant images from multiple view-
TABLE IV
CLA SSI FICATI ON P ERF ORM ANC E OF VARI OU S DEE P LEA RN ING M ODE LS
ON T HE CWD30 DATASE T,COM PARI NG RE SULT S OBTAI NE D FROM
RA NDO M INI TIA LI ZATIO N AND IM AGE NET IN IT IAL IZATI ON .
Typ. Methods Scratch ImageNet-1K
F1 Acc F1 Acc
CNN
ResNet-101 [55] 76.38 80.17 83.83 88.66
ResNext-101 [56] 79.76 81.36 84.03 89.06
MobileNetv3-L [57] 74.67 78.95 81.80 86.29
EfficientNetv2-M [58] 87.37 83.06 84.91 90.79
Trans.
ViT [59] 78.90 83.43 84.08 87.84
SwinViT [60] 81.53 87.59 83.70 88.71
MaxViT [61] 82.24 87.08 82.43 91.45

9
Fig. 9. Illustration how simple image processing techniques can transform CWD30 dataset into related subsets, emphasizing CWD30 as a comprehensive
superset.
ing angles and at different growth stages, can accelerate our
understanding of genotype-phenotype relationships. It can also
assist plant scientists and breeders in developing advanced
phenotyping systems that offer more detailed phenotypic in-
formation about plants. Figure 7a displays randomly selected
samples of crops and weeds at different life cycle stages, with
images captured at a 90-degree angle from the plant. The
graph in Figure 7b show the distribution of images across
each growing stage.
F. Comparison with Other Datasets
In this section, we compare the CWD30 dataset with several
existing datasets related to crop-weed recognition. Our dataset
stands out as a more holistic, domain-adverse, versatile, and
diverse dataset that provides a comprehensive solution to
crop-weed discrimination. Furthermore, it classifies weeds into
major families, such as grasses, sedges, and broad leaves, and
further into specific weed sub-categories. To the best of our
Fig. 10. Graph comparing deep learning models in terms of parameters (in
million), feature embeddings (no. of features), and forward and backward
pass sizes (in megabytes), highlighting the trade-offs among the models. Best
viewed in color.
knowledge, CWD30 is the first dataset of its kind in the field
of practical crop-weed discrimination.
The PDD271 dataset contains close-up images of only dis-
eased plant parts, the Deep Weeds dataset has low-resolution
images of roadside weeds, and the Plant Seedling dataset
consists of early-stage weeds grown in lab trays. The most
comparable dataset in this field is the CNU weed dataset,
which focuses on field environments but features simplified
representations of plants, i.e., zoomed in part of plants.
Existing data sets’ shortcomings can be summarized as
follows:
1) Simplified representation: By focusing on specific plant
parts, such as leaves or branches, the data becomes less
complex and fails to represent real-field challenges.
2) Limited scope: Images of specific plant parts may not
capture the full characteristics of a plant, leading to less
accurate recognition systems.
3) Restricted environments: Capturing images in specific
fields may limit the model’s ability to generalize to other
settings or conditions.
4) Less robust features: The absence of multiple angles and
growth stages may result in less robust feature learning
and hinder the model’s ability to handle occlusions,
rotations, and scale variations.
5) Smaller dataset size: Most existing precision agricultural
datasets have a limited number of images, hindering
the development of more advanced deep learning-based
systems.
In contrast, the CWD30 dataset addresses these limitations
with the following inherent properties:
1) Comprehensive representation: Full-plant images pro-
vide a holistic view, capturing various aspects of the
crops and weeds.
2) Varied environments: Capturing plants in both indoor
and outdoor settings enable the dataset to cover a broader
range of conditions and will enhance the model’s gen-
eralizability.

10
TABLE V
PER FOR MAN CE CO MPAR ISO N OF D EEP L EAR NI NG MO DEL S USI NG P RET RAI NED W EI GHT S FRO M IMAGE NET AN D CWD30, HI GHL IGH TI NG TH E IMPAC T
OF DATAS ET-SPE CI FIC PR ETR AIN IN G ON MO DEL P ERF OR MAN CE.
Typ. Methods Deep Weeds [25] Plant Seedlings [26] Cassava Plant [34] IP 102 [38]
ImageNet-1k CWD-30 ImageNet-1k CWD-30 ImageNet-1k CWD-30 ImageNet-1k CWD-30
CNN
ResNet-101 [55] 91.13 95.08 90.14 96.27 64.82 71.44 60.34 66.87
ResNext-101 [56] 90.70 95.87 92.46 97.79 65.01 73.22 62.13 67.90
MobileNetv3-L [57] 89.08 94.62 88.43 96.54 66.34 71.17 61.08 64.53
EfficientNetv2-M [58] 91.39 95.78 90.85 97.18 61.13 69.34 60.86 68.29
Trans.
ViT [59] 86.25 90.18 91.41 95.39 58.24 61.32 59.77 68.46
SwinViT [60] 88.83 96.70 93.24 98.06 73.83 78.66 59.11 68.67
MaxViT [61] 87.79 97.04 92.47 97.89 71.55 79.54 60.51 69.36
3) Multiple angles: Images taken from different angles
allow models to learn robust features and improve occlu-
sion handling, rotation invariance, and scale invariance.
4) Different growth stages: Capturing images at various
growth stages helps models recognize crops and weeds
at any stage of their life cycle, resulting in more accurate
and reliable CAPA systems.
5) Complexity: Increased variability and complexity make
the images more challenging to analyze.
6) Larger dataset size: The proposed dataset is one of
the largest real-image datasets to date in the field of
precision agriculture.
By addressing domain-specific challenges in real-field agri-
cultural environments and providing a diverse, varied, and
extensive collection of images, CWD30 advances research in
the field and enhances data efficiency and performance in a
wide range of downstream agricultural tasks.
An additional advantage of the CWD30 dataset is its ver-
satility, which allows it to encompass various existing agri-
cultural datasets through simple image processing operations.
By applying random cropping, downsampling, foreground
segmentation, or thresholding to the images in the CWD30
dataset, one can create subsets that resemble other datasets in
the field. An example of this process is shown in Figure 8.
This demonstrates that the CWD30 dataset can be considered
a comprehensive and unified source of agricultural data, with
other datasets effectively serving as subsets of CWD30. This
versatility not only highlights the extensive nature of the
CWD30 dataset but also supports its potential for advancing
research and improving performance in a wide range of
agricultural tasks.
G. Data Imbalance Ration
A dataset’s imbalance ratio (IR) refers to the degree of
disparity between the number of samples in different classes
[62]. In the context of deep learning, the imbalance ratio can
have significant effects on model performance. Although low
data imbalance ratios in datasets, like MNIST and ImageNet-
1K, are generally preferred for deep learning models as they
promote balanced class representation and accurate perfor-
mance, these datasets do not always represent real-world
situations where data samples for some classes are harder to
obtain.
In contrast, high data imbalance ratios, found in datasets
such as CNU, CWD30, and DeepWeeds, can pose challenges
for deep learning models as they may lead to overfitting
and poor generalization. Models trained on highly imbalanced
datasets can become biased towards majority classes, result-
ing in decreased performance for minority classes. However,
one key advantage of having high imbalance ratios is their
increased representation of real-world situations, particularly
in complex recognition tasks like precision agriculture, where
some classes naturally have fewer available samples. While
these imbalanced datasets present challenges, they also offer
a more realistic depiction of real-world scenarios, pushing
deep learning models to adapt and improve their performance
in diverse and unevenly distributed data conditions. Figure 9
shows imbalance ration of related datasets.
To the best of our knowledge, the proposed CWD30
dataset offers several distinctive features not found in previous
datasets, as highlighted in earlier sub-sections. These features
can bridge the genotyping-phenotyping gap, enhance the ro-
bustness and reliability of deep learning systems, and expand
their area of applications.
IV. EXP ER IM EN TS AND EVALUATION
We conducted a comprehensive experimental evaluation of
the CWD30 dataset, focusing on classification performance
using deep convolutional and transformer-based architectures.
Additionally, we examine the influence of CWD30 pretrained
networks on downstream precision agriculture tasks, including
semantic segmentation.
A. Experimental Setup
In our experiments all networks’ layers are fine-tuned using
an AdamW optimizer with a minibatch size of 32 and an
initial learning rate of 6e-5. We employ a cosine decay
policy for reducing the learning rate and incorporate a dropout
value of 0.2, along with basic data augmentations, to prevent
overfitting. While the deep models’ fundamental architectures
remain unchanged, the last fully connected layer is adapted
to match the number of target classification classes. Each
network is trained for 50 epochs across all datasets, and the
reported results represent the average of three runs. Input
images are resized to 224 x 224 pixels. Our deep feature-based
experiments are implemented using PyTorch and performed
on an NVIDIA Titan RTX-3090 GPU with 24 GB of onboard
memory.

11
Fig. 11. 2D t-SNE feature embeddings visualization comparing best performing deep learning model (i.e., MaxViT) with pretrained weights from ImageNet
and CWD30, on various agricultural datasets. Highlighting the improved cluster patterns and separation achieved using the CWD30 pretrained network.
B. Evaluation Metrics
To objectively assess models trained on the CWD30 dataset,
we employ widely accepted evaluation metrics for comprehen-
sive comparisons. Given the dataset’s imbalanced class distri-
bution, we utilize the following metrics for better performance
assessment:
•Per-class Mean Accuracy (Acc) calculates the average
of individual class mean accuracies, providing a bal-
anced performance evaluation, especially for imbalanced
datasets like CWD30.
•F1-Score is the harmonic mean of precision (the ratio of
true positive predictions to the sum of true positive and
false positive predictions) and recall (the ratio of true
positive predictions to the sum of true positive and false
negative predictions), offering a single value representing
the model’s overall performance while accounting for
false positive and false negative errors.
•For downstream tasks like semantic segmentation, we use
mean intersection over union (mIoU), which evaluates the
overlap between predicted and ground truth segments.
By examining these metrics, researchers can identify the
most promising approaches to guide future developments in
precision agriculture and the development of CAPA systems.
V. RE SU LTS AN D DIS CU SS ION
In this section, we present the classification results for
various deep learning models trained on the CWD30 dataset.
We compare the models [55]–[61] based on their F1-Score
and per-class mean accuracy (Acc) when trained from scratch
and when pretrained on the ImageNet-1K dataset. The re-
sults are summarized in the table IV. The results reveal
that EfficientNetv2-M [58] is the best-performing CNN ar-
chitecture when trained from scratch, with the highest F1-
Score (82.37) and accuracy (87.06). Pretraining on ImageNet-
1K consistently improves the performance of all models.
Among transformer-based models, SwinViT [60] achieves
the highest accuracy (88.71), and MaxViT [61] obtains the
highest F1-Score (82.43). Generally, more complex models
like EfficientNetv2-M and MaxViT outperform less complex
counterparts, as their increased capacity better captures and
represents the nuances in the CWD30 dataset.
Moreover, transformer-based models like SwinViT and
MaxViT demonstrate superior performance compared to their
CNN counterparts despite having fewer parameters and a
smaller memory footprint (forward and backward pass). This
observation underscores the potential of transformer architec-
tures for handling the diverse and complex patterns in the
CWD30 dataset. The self-attention mechanism in transformers
may allow them to capture long-range dependencies and fine-
grained patterns more effectively than traditional convolutional
layers.
Additionally, we compare the model parameters and mem-
ory footprint against the final output feature embeddings gen-
erated by the model just before the linear classification layers,
as shown in the figure 9. Intriguingly, MaxViT, which outputs
the fewest feature embeddings (512), still outperforms all other
models. This finding is significant because lower-dimensional
feature embeddings offer practical advantages for real-world
applications, especially in resource-constrained environments.
For instance, in precision agriculture, heavy GPUs like the
RTX-3090 may not be suitable for field deployment due
to their large size and power consumption. Instead, smaller
embedded systems like NVIDIA Jetson boards are com-
monly used, which have limited memory and computational
resources. By employing deep learning models with lower-
dimensional embeddings, parameters, and memory footprint,
these systems can efficiently process and analyze data, making
them more suitable for real-world applications.

12
Fig. 12. Sample images from (a) SugarBeet [63], (b) BeanWeed [64] and (c)
CarrotWeed [65] datasets.
The diverse and sizable CWD30 dataset is essential for
the development of robust and reliable CAPA systems, as it
offers a rich source of real-world precision agriculture data for
training deep data hungry models. By focusing on the quality
of the dataset and addressing practical constraints of real-world
deployments, researchers can ensure that deep learning models
are capable of handling inherent variability and imbalances in
agricultural settings, ultimately making them more efficient,
generalizable, and suitable for a wide range of applications,
including field deployment.
A. Further Analysis
To further evaluate the performance enhancements offered
by using the CWD30 dataset for pretraining and finetuning
on tasks with limited samples, we tested multiple publicly
available benchmark agricultural datasets [25], [26], [34],
[38] for robust feature extraction and compared the results
with models pretrained on the ImageNet-1K dataset. Detailed
information about these datasets is provided in Section II. For
each dataset, we adhere to the testing and data split settings
outlined in their original papers, while maintaining the same
network training settings as described in the previous subsec-
tion. The results are summarized in Table V. Throughout all
datasets MaxViT achieved highest per class mean accuracy
scores despite having minimum output feature embeddings.
Whereas pretraining on CWD30 dataset consistently improves
the performance of all tested architectures on all datasets.
For better understanding and comparison, we extract high-
dimensional feature embeddings (features of second last layer)
from the best-performing model, i.e., MaxViT, on test images
of all datasets. The compactness and expressiveness of these
feature embeddings facilitate the development of efficient
and accurate algorithms for various applications, including
CAPA systems. We perform t-SNE [66] visualization on
these feature embeddings. t-SNE, effectively projects high-
dimensional feature embeddings onto a two-dimensional space
while preserving the local structure and relationships within
the data. By plotting t-SNE visualizations, we can assess the
separability and distribution of the data in the reduced space,
as well as the quality of the learned feature representations.
Our results reveal that models pretrained on the CWD30
dataset produce more distinct and well-separated clusters in
the t-SNE plots when fine-tuned on various public datasets
compared to ImageNet pretrained models. The t-SNE plots
for CWD30 and ImageNet pretrained MaxViT models on
publicly available datasets are displayed in Figure 11. From
the Figure 11, it is evident that CWD30-pretrained models
learn more meaningful and robust feature representations, as
the clusters in these plots are better defined and distinct, with
points belonging to the same cluster positioned closer together
and clear separation between clusters. This ultimately leads
to improved performance during finetuning and downstream
tasks (see section V.B).
B. Performance on Downstream Tasks
To evaluate the effectiveness of enhanced feature repre-
sentations obtained by CWD30 pretraining on downstream
tasks, we assess several state-of-the-art segmentation models
for pixel-level crop weed recognition. We use three publicly
available crop-weed datasets: CarrotWeed [65], SugarBeet
[63], and BeanWeed [64]. Sample images from each dataset,
along with their corresponding segmentation labels, are shown
in Figure. The quantitative results are summarized in Table VI.
Throughout the experiments, it is evident that pretraining ar-
chitecture backbones with CWD30 provides a clear advantage
over ImageNet-1K pretrained backbones. Although the perfor-
mance difference may not appear substantial when examining
the table VI, the difference becomes more apparent when ana-
lyzing the learning curves of both setups. The learning curves
of the best-performing SegNext [70] model are shown in Fig-
ure 13. These curves demonstrate that initializing experiments
with weights obtained from training on more relevant datasets
(i.e., agricultural data) results in faster convergence and stable
training. From the plots, it can be seen that the difference
between ImageNet and CWD30 initialization is significant at
the 10th epoch, where the CWD30-initialized model already
reaches performance close to its final convergence value. In
contrast, for ImageNet initialized models, it takes about 50
epochs to achieve similar performance.
These findings in this section underscore the importance of
employing a comprehensive agricultural dataset like CWD30
for pretraining deep learning models. By utilizing the rich
and diverse data offered by CWD30, researchers can develop
efficient and generalizable deep learning models that are more
suitable for a wide range of applications, including precision
agriculture.
VI. CO NCLUS IO N
In conclusion, this paper presents the CWD30 dataset, a
comprehensive, holistic, large-scale, and diverse crop-weed
recognition dataset tailored for precision agriculture. With over
219,770 high-resolution images of 20 weed species and 10
crop species, the dataset spans various growth stages, multiple
viewing angles, and diverse environmental conditions. The
hierarchical taxonomy of CWD30 facilitates the development
of accurate, robust, and generalizable deep learning models
for crop-weed recognition. Our extensive baseline experiments
demonstrate the challenges and opportunities presented by

13
TABLE VI
COM PARIS ON O F PER FOR MAN CE O N DOWN ST REA M SEG ME NTATIO N TASKS U SI NG PR ETR AIN ED BAC KB ONE S (I.E., IM AGE NET VS . CWD30).
Method Backbone SugarBeet CarrotWeed BeanWeed
ImageNet-1k CWD-30 ImageNet-1k CWD-30 ImageNet-1k CWD-30
U-Net [67] ResNet-101 [55] 80.96 85.47 75.47 78.32 69.67 72.49
DeepLabv3+ [68] ResNet-101 [55] 81.17 86.02 80.29 83.16 72.41 78.03
OCR [69] ResNet-101 [55] 84.79 87.34 84.56 86.53 73.60 79.51
SegNeXt-L [70] MSCAN [70] 84.15 87.65 83.79 88.54 80.05 83.90
Fig. 13. Learning curves illustrating the superior performance and faster convergence of CWD30 pretrained backbones on downstream segmentation tasks.(a)
SugarBeet [63], (b) CarrotWeed [65] and (c) BeanWeed [64].
the CWD30 dataset. These experiments emphasize the im-
portance of utilizing CWD30 pretrained backbones, which
result in enhanced performance, reduced convergence time,
and consequently, saved time and training resources for various
fine-tuning and downstream precision agriculture tasks. The
CWD30 dataset not only advances research in the field of
precision agriculture but also promotes collaboration among
researchers by serving as a benchmark for evaluating crop-
weed recognition algorithms.
ACK NOW LE DG MENTS
This work was supported in part by the Agricultural Science
and Technology Development Cooperation Research Program
(PJ015720) and Basic Science Research Program through the
National Research Foundation of Korea (NRF) funded by
the Ministry of Education (NRF-2019R1A6A1A09031717 and
NRF-2019R1A2C1011297).
APP ENDIX
TAXO NO MY O F PLANT SPE CI ES
See Table VII.
REF ERENCES
[1] P. Radoglou-Grammatikis, P. Sarigiannidis, T. Lagkas, and I. Moscho-
lios, “A compilation of uav applications for precision agriculture,”
Computer Networks, vol. 172, p. 107148, 2020.
[2] N. Iqbal, S. Manalil, B. S. Chauhan, and S. W. Adkins, “Investigation
of alternate herbicides for effective weed management in glyphosate-
tolerant cotton,” Archives of Agronomy and Soil Science, vol. 65, no. 13,
pp. 1885–1899, 2019.
[3] D. Patel and B. Kumbhar, “Weed and its management: A major threats
to crop economy,” Journal Pharmaceutical Science and Bioscientific
Research (JPSBR), vol. 6, no. 6, pp. 753–758, 2016.
[4] S. I. Moazzam, U. S. Khan, M. I. Tiwana, J. Iqbal, W. S. Qureshi, and
S. I. Shah, “A review of application of deep learning for weeds and
crops classification in agriculture,” in 2019 International Conference on
Robotics and Automation in Industry (ICRAI). IEEE, 2019, pp. 1–6.
[5] T. Ilyas, H. Jin, M. I. Siddique, S. J. Lee, H. Kim, and L. Chua, “Diana:
A deep learning-based paprika plant disease and pest phenotyping
system with disease severity analysis,” Frontiers in Plant Science, p.
3862, 2022.
[6] O. Elsherbiny, L. Zhou, L. Feng, and Z. Qiu, “Integration of visible and
thermal imagery with an artificial neural network approach for robust
forecasting of canopy water content in rice,” Remote Sensing, vol. 13,
no. 9, p. 1785, 2021.
[7] I. Sa, Z. Chen, M. Popovi´
c, R. Khanna, F. Liebisch, J. Nieto, and R. Sieg-
wart, “weednet: Dense semantic weed classification using multispectral
images and mav for smart farming,” IEEE robotics and automation
letters, vol. 3, no. 1, pp. 588–595, 2017.
[8] A. M. Hasan, F. Sohel, D. Diepeveen, H. Laga, and M. G. Jones, “A
survey of deep learning techniques for weed detection from images,”
Computers and Electronics in Agriculture, vol. 184, p. 106067, 2021.
[9] Y. Bai, J. Mei, A. L. Yuille, and C. Xie, “Are transformers more
robust than cnns?” Advances in Neural Information Processing Systems,
vol. 34, pp. 26 831–26 843, 2021.
[10] A. Joshi, D. Guevara, and M. Earles, “Standardizing and centralizing
datasets to enable efficient training of agricultural deep learning models,”
arXiv preprint arXiv:2208.02707, 2022.
[11] C. Shorten and T. M. Khoshgoftaar, “A survey on image data augmen-
tation for deep learning,” Journal of big data, vol. 6, no. 1, pp. 1–48,
2019.
[12] D. Su, H. Kong, Y. Qiao, and S. Sukkarieh, “Data augmentation for
deep learning based semantic segmentation and crop-weed classification
in agricultural robotics,” Computers and Electronics in Agriculture, vol.
190, p. 106418, 2021.
[13] C. Shorten and T. M. Khoshgoftaar, “A survey on image data augmen-
tation for deep learning,” Journal of big data, vol. 6, no. 1, pp. 1–48,
2019.
[14] B. Espejo-Garcia, N. Mylonas, L. Athanasakos, S. Fountas, and I. Vasi-
lakoglou, “Towards weeds identification assistance through transfer
learning,” Computers and Electronics in Agriculture, vol. 171, p.
105306, 2020.
[15] Q. H. Cap, H. Uga, S. Kagiwada, and H. Iyatomi, “Leafgan: An effective
data augmentation method for practical plant disease diagnosis,” IEEE
Transactions on Automation Science and Engineering, vol. 19, no. 2,
pp. 1258–1267, 2020.

14
TABLE VII
DETAI LED TA XON OMY O F PLAN T SPE CIE S INCL UD ED IN T HE CW D30 DATAS ET. TH E KIN GD OM AN D PHY LUM O F AL L PLA NTS L IST ED A RE PL ANTA E
AN D MAGN OLI OP HYTA R ESP ECT IVE LY.
Common Name Scientific Name Order Family Genus Species Class Sub-Class
Asian flatsedge Cyperus microiria Poales Cyperaceae Cyperus microiria Weed broad-leaves
Asiatic dayflower Commelina communis Commelinales Commelinaceae Commelina communis Weed broad-leaves
Bean Phaseolus vulgaris Fabales Fabaceae Phaseolus vulgaris Crop legumes
Bloodscale sedge Carex haematostoma Poales Cyperaceae Carex haematostoma Weed sedge
Cockspur grass Echinochloa crus-galli Poales Poaceae Echinochloa crus-galli Weed grass
Copperleaf Acalypha spp. Malpighiales Euphorbiaceae Acalypha spp. Weed broad-leaves
Corn Zea mays Poales Poaceae Zea mays Crop grains
Early barnyard grass Echinochloa oryzoides Poales Poaceae Echinochloa oryzoides Weed grass
Fall panicum Panicum dichotomiflorum Poales Poaceae Panicum dichotomiflorum Weed grass
Finger grass Digitaria sanguinalis Poales Poaceae Digitaria sanguinalis Weed grass
Foxtail millet Setaria italica Poales Poaceae Setaria italica Crop grains
Goosefoot Chenopodium album Caryophyllales Amaranthaceae Chenopodium album Weed broad-leaves
Great millet Sorghum bicolor Poales Poaceae Sorghum bicolor Crop grains
Green foxtail Setaria viridis Poales Poaceae Setaria viridis Weed grass
Green gram Vigna radiata Fabales Fabaceae Vigna radiata Crop legumes
Henbit Lamium amplexicaule Lamiales Lamiaceae Lamium amplexicaule Weed broad-leaves
Indian goosegrass Eleusine indica Poales Poaceae Eleusine indica Weed grass
Korean dock Rumex crispus Caryophyllales Polygonaceae Rumex crispus Weed broad-leaves
Livid pigweed Amaranthus lividus Caryophyllales Amaranthaceae Amaranthus lividus Weed broad-leaves
Nipponicus sedge Carex nipponica Poales Cyperaceae Carex nipponica Weed sedge
Peanut Arachis hypogaea Fabales Fabaceae Arachis hypogaea Crop broad-leaves
Perilla Perilla frutescens Lamiales Lamiaceae Perilla frutescens Crop oil seeds
Poa annua Poa annua Poales Poaceae Poa annua Weed grasses
Proso millet Panicum miliaceum Poales Poaceae Panicum miliaceum Crop grains
Purslane Portulaca oleracea Caryophyllales Portulacaceae Portulaca oleracea Weed broad-leaves
Red bean Phaseolus angularis Fabales Fabaceae Phaseolus angularis Crop broad-leaves
Redroot pigweed Amaranthus retroflexus Caryophyllales Amaranthaceae Amaranthus retroflexus Weed broad-leaves
Sesame Sesamum indicum Lamiales Pedaliaceae Sesamum indicum Crop oil seeds
Smooth pigweed Amaranthus hybridus Caryophyllales Amaranthaceae Amaranthus hybridus Weed broad-leaves
White goosefoot Chenopodium album Caryophyllales Amaranthaceae Chenopodium album Weed broad-leaves
[16] T. Moon and J. E. Son, “Knowledge transfer for adapting pre-trained
deep neural models to predict different greenhouse environments based
on a low quantity of data,” Computers and Electronics in Agriculture,
vol. 185, p. 106136, 2021.
[17] S. J. Pan and Q. Yang, “A survey on transfer learning,” IEEE Trans-
actions on knowledge and data engineering, vol. 22, no. 10, pp. 1345–
1359, 2010.
[18] O. Antonijevi´
c, S. Jeli´
c, B. Bajat, and M. Kilibarda, “Transfer learning
approach based on satellite image time series for the crop classification
problem,” Journal of Big Data, vol. 10, no. 1, pp. 1–19, 2023.
[19] J. Deng, W. Dong, R. Socher, L.-J. Li, K. Li, and L. Fei-Fei, “Imagenet:
A large-scale hierarchical image database,” in 2009 IEEE conference on
computer vision and pattern recognition. Ieee, 2009, pp. 248–255.
[20] T.-Y. Lin, M. Maire, S. Belongie, J. Hays, P. Perona, D. Ramanan,
P. Doll´
ar, and C. L. Zitnick, “Microsoft coco: Common objects in
context,” in Computer Vision–ECCV 2014: 13th European Conference,
Zurich, Switzerland, September 6-12, 2014, Proceedings, Part V 13.
Springer, 2014, pp. 740–755.
[21] Z. Al Sahili and M. Awad, “The power of transfer learning in agricultural
applications: Agrinet,” Convolutional neural networks and deep learning
for crop improvement and production, vol. 16648714, p. 195, 2023.
[22] A. Kamilaris and F. X. Prenafeta-Bold´
u, “Deep learning in agriculture: A
survey,” Computers and electronics in agriculture, vol. 147, pp. 70–90,
2018.
[23] A. Fuentes, S. Yoon, and D. S. Park, “Deep learning-based phenotyping
system with glocal description of plant anomalies and symptoms,”
Frontiers in Plant Science, vol. 10, p. 1321, 2019.
[24] M. Rahnemoonfar and C. Sheppard, “Deep count: fruit counting based
on deep simulated learning,” Sensors, vol. 17, no. 4, p. 905, 2017.
[25] A. Olsen, D. A. Konovalov, B. Philippa, P. Ridd, J. C. Wood, J. Johns,
W. Banks, B. Girgenti, O. Kenny, J. Whinney et al., “Deepweeds: A
multiclass weed species image dataset for deep learning,” Scientific
reports, vol. 9, no. 1, p. 2058, 2019.
[26] T. M. Giselsson, R. N. Jørgensen, P. K. Jensen, M. Dyrmann, and H. S.
Midtiby, “A public image database for benchmark of plant seedling
classification algorithms,” arXiv preprint arXiv:1711.05458, 2017.
[27] S. S. Chouhan, U. P. Singh, A. Kaul, and S. Jain, “A data repository of
leaf images: Practice towards plant conservation with plant pathology,”
in 2019 4th International Conference on Information Systems and
Computer Networks (ISCON). IEEE, 2019, pp. 700–707.
[28] J. G. A. Barbedo, “Plant disease identification from individual lesions
and spots using deep learning,” Biosystems Engineering, vol. 180, pp.
96–107, 2019.
[29] X. Qian, C. Zhang, L. Chen, and K. Li, “Deep learning-based identifi-
cation of maize leaf diseases is improved by an attention mechanism:
Self-attention,” Frontiers in Plant Science, p. 1154, 2022.
[30] L. Goyal, C. M. Sharma, A. Singh, and P. K. Singh, “Leaf and
spike wheat disease detection & classification using an improved deep
convolutional architecture,” Informatics in Medicine Unlocked, vol. 25,
p. 100642, 2021.
[31] D. Hughes, M. Salath´
eet al., “An open access repository of images on
plant health to enable the development of mobile disease diagnostics,”
arXiv preprint arXiv:1511.08060, 2015.
[32] D. Singh, N. Jain, P. Jain, P. Kayal, S. Kumawat, and N. Batra,
“Plantdoc: A dataset for visual plant disease detection,” in Proceedings
of the 7th ACM IKDD CoDS and 25th COMAD, 2020, pp. 249–253.
[33] P. K. Sethy, N. K. Barpanda, A. K. Rath, and S. K. Behera, “Deep feature
based rice leaf disease identification using support vector machine,”
Computers and Electronics in Agriculture, vol. 175, p. 105527, 2020.
[34] H. Ayu, A. Surtono, and D. Apriyanto, “Deep learning for detection
cassava leaf disease,” in Journal of Physics: Conference Series, vol.
1751, no. 1. IOP Publishing, 2021, p. 012072.
[35] R. Thapa, N. Snavely, S. Belongie, and A. Khan, “The plant pathology
2020 challenge dataset to classify foliar disease of apples,” arXiv
preprint arXiv:2004.11958, 2020.
[36] V. H. Trong, Y. Gwang-hyun, D. T. Vu, and K. Jin-young, “Late fusion of
multimodal deep neural networks for weeds classification,” Computers
and Electronics in Agriculture, vol. 175, p. 105506, 2020.
[37] X. Liu, W. Min, S. Mei, L. Wang, and S. Jiang, “Plant disease
recognition: A large-scale benchmark dataset and a visual region and
loss reweighting approach,” IEEE Transactions on Image Processing,
vol. 30, pp. 2003–2015, 2021.
[38] X. Wu, C. Zhan, Y.-K. Lai, M.-M. Cheng, and J. Yang, “Ip102: A large-
scale benchmark dataset for insect pest recognition,” in Proceedings of
the IEEE/CVF conference on computer vision and pattern recognition,
2019, pp. 8787–8796.

15
[39] D. M. S. Arsa, T. Ilyas, S.-H. Park, O. Won, and H. Kim, “Eco-friendly
weeding through precise detection of growing points via efficient multi-
branch convolutional neural networks,” Computers and Electronics in
Agriculture, vol. 209, p. 107830, 2023.
[40] H. Escalante, S. Rodr´
ıguez-S´
anchez, M. Jim´
enez-Liz´
arraga, A. Morales-
Reyes, J. De La Calleja, and R. Vazquez, “Barley yield and fertilization
analysis from uav imagery: a deep learning approach,” International
journal of remote sensing, vol. 40, no. 7, pp. 2493–2516, 2019.
[41] J. Yi, L. Krusenbaum, P. Unger, H. H ¨
uging, S. J. Seidel, G. Schaaf, and
J. Gall, “Deep learning for non-invasive diagnosis of nutrient deficiencies
in sugar beet using rgb images,” Sensors, vol. 20, no. 20, p. 5893, 2020.
[42] J. H. Westwood, R. Charudattan, S. O. Duke, S. A. Fennimore, P. Mar-
rone, D. C. Slaughter, C. Swanton, and R. Zollinger, “Weed management
in 2050: Perspectives on the future of weed science,” Weed science,
vol. 66, no. 3, pp. 275–285, 2018.
[43] A. Wang, W. Zhang, and X. Wei, “A review on weed detection
using ground-based machine vision and image processing techniques,”
Computers and electronics in agriculture, vol. 158, pp. 226–240, 2019.
[44] A. Khan, A. D. Vibhute, S. Mali, and C. Patil, “A systematic review
on hyperspectral imaging technology with a machine and deep learning
methodology for agricultural applications,” Ecological Informatics, p.
101678, 2022.
[45] G. R. Coleman and W. Salter, “More eyes on the prize: open-source
data, software and hardware for advancing plant science through collab-
oration,” AoB Plants, p. plad010, 2023.
[46] “Plants database,” 5 2023, last Accessed: 2023-05-09. [Online].
Available: https://plants.usda.gov/home/raritySearch
[47] “Weed surveys,” 5 2023, last Accessed: 2023-05-09. [Online]. Available:
https://wssa.net/wssa/weed/surveys/
[48] “Agriculture production data,” 4 2023, last Accessed: 2023-05-09.
[Online]. Available: http://www.rda.go.kr/board/board.do?mode=html&
prgId=oda opendata
[49] “World agricultural production,” 4 2023, last Accessed: 2023-05-
09. [Online]. Available: https://apps.fas.usda.gov/psdonline/circulars/
production.pdf
[50] B. Leff, N. Ramankutty, and J. A. Foley, “Geographic distribution of
major crops across the world,” Global biogeochemical cycles, vol. 18,
no. 1, 2004.
[51] Y. Lu and S. Young, “A survey of public datasets for computer
vision tasks in precision agriculture,” Computers and Electronics in
Agriculture, vol. 178, p. 105760, 2020.
[52] W. Coudron, A. Gobin, C. Boeckaert, T. De Cuypere, P. Lootens,
S. Pollet, K. Verheyen, P. De Frenne, and T. De Swaef, “Data collection
design for calibration of crop models using practical identifiability
analysis,” Computers and Electronics in Agriculture, vol. 190, p. 106457,
2021.
[53] W. Coudron, P. De Frenne, K. Verheyen, A. Gobin, C. Boeckaert,
T. De Cuypere, P. Lootens, S. Pollet, and T. De Swaef, “Usefulness
of cultivar-level calibration of aquacrop for vegetables depends on the
crop and data availability,” Frontiers in Plant Science, vol. 14, 2023.
[54] S. Yadav and S. Shukla, “Analysis of k-fold cross-validation over hold-
out validation on colossal datasets for quality classification,” in 2016
IEEE 6th International conference on advanced computing (IACC).
IEEE, 2016, pp. 78–83.
[55] K. He, X. Zhang, S. Ren, and J. Sun, “Deep residual learning for image
recognition,” in Proceedings of the IEEE conference on computer vision
and pattern recognition, 2016, pp. 770–778.
[56] S. Xie, R. Girshick, P. Doll´
ar, Z. Tu, and K. He, “Aggregated residual
transformations for deep neural networks,” in Proceedings of the IEEE
conference on computer vision and pattern recognition, 2017, pp. 1492–
1500.
[57] A. Howard, M. Sandler, G. Chu, L.-C. Chen, B. Chen, M. Tan, W. Wang,
Y. Zhu, R. Pang, V. Vasudevan et al., “Searching for mobilenetv3,”
in Proceedings of the IEEE/CVF international conference on computer
vision, 2019, pp. 1314–1324.
[58] M. Tan and Q. Le, “Efficientnetv2: Smaller models and faster training,”
in International conference on machine learning. PMLR, 2021, pp.
10 096–10 106.
[59] A. Dosovitskiy, L. Beyer, A. Kolesnikov, D. Weissenborn, X. Zhai,
T. Unterthiner, M. Dehghani, M. Minderer, G. Heigold, S. Gelly et al.,
“An image is worth 16x16 words: Transformers for image recognition
at scale,” arXiv preprint arXiv:2010.11929, 2020.
[60] Z. Liu, Y. Lin, Y. Cao, H. Hu, Y. Wei, Z. Zhang, S. Lin, and
B. Guo, “Swin transformer: Hierarchical vision transformer using shifted
windows,” in Proceedings of the IEEE/CVF international conference on
computer vision, 2021, pp. 10 012–10 022.
[61] Z. Tu, H. Talebi, H. Zhang, F. Yang, P. Milanfar, A. Bovik, and Y. Li,
“Maxvit: Multi-axis vision transformer,” in Computer Vision–ECCV
2022: 17th European Conference, Tel Aviv, Israel, October 23–27, 2022,
Proceedings, Part XXIV. Springer, 2022, pp. 459–479.
[62] J. M. Johnson and T. M. Khoshgoftaar, “Survey on deep learning with
class imbalance,” Journal of Big Data, vol. 6, no. 1, pp. 1–54, 2019.
[63] N. Chebrolu, P. Lottes, A. Schaefer, W. Winterhalter, W. Burgard,
and C. Stachniss, “Agricultural robot dataset for plant classification,
localization and mapping on sugar beet fields,” The International Journal
of Robotics Research, 2017.
[64] T. Ilyas, H. Kim, J. Lee, O. Won, and Y. Jeong, “Adaptive deep learning
for crop weed discrimination in unseen fields,” Available at SSRN
4345158, 2023.
[65] S. Haug and J. Ostermann, “A crop/weed field image dataset for
the evaluation of computer vision based precision agriculture tasks,”
in Computer Vision - ECCV 2014 Workshops, 2015, pp. 105–116.
[Online]. Available: http://dx.doi.org/10.1007/978-3- 319-16220- 1 8
[66] T. T. Cai and R. Ma, “Theoretical foundations of t-sne for visualizing
high-dimensional clustered data,” The Journal of Machine Learning
Research, vol. 23, no. 1, pp. 13581–13 634, 2022.
[67] O. Ronneberger, P. Fischer, and T. Brox, “U-net: Convolutional networks
for biomedical image segmentation,” in Medical Image Computing
and Computer-Assisted Intervention–MICCAI 2015: 18th International
Conference, Munich, Germany, October 5-9, 2015, Proceedings, Part III
18. Springer, 2015, pp. 234–241.
[68] L.-C. Chen, G. Papandreou, F. Schroff, and H. Adam, “Rethinking
atrous convolution for semantic image segmentation,” arXiv preprint
arXiv:1706.05587, 2017.
[69] Y. Yuan, X. Chen, and J. Wang, “Object-contextual representations
for semantic segmentation,” in Computer Vision–ECCV 2020: 16th
European Conference, Glasgow, UK, August 23–28, 2020, Proceedings,
Part VI 16. Springer, 2020, pp. 173–190.
[70] M.-H. Guo, C.-Z. Lu, Q. Hou, Z. Liu, M.-M. Cheng, and S.-M.
Hu, “Segnext: Rethinking convolutional attention design for semantic
segmentation,” arXiv preprint arXiv:2209.08575, 2022.