Download as pdf or txt
Download as pdf or txt
You are on page 1of 11

Comparative Analysis of Traditional classification

with Hybrid Segmentation Approaches and deep


learning classification in Computer Vision
1st Reza Wahyu Ramadhan
Department of Informatics
Institut Teknologi Sepuluh Nopember
Surabaya, Indonesia

Abstract—This research conducted comparing hybrid segmen- images of blood cells. This method not only improves the
tation using K-Nearest Neighbor with Nick Thresholding applied classification effectiveness but also reduces computational time
to Naive Bayes as traditional classification and a new convo- significantly, achieving a notable accuracy rate of 98.87% with
lutional neural network named KR-NET in computer vision.
Specifically the research tried focusing on examining the perfor- the integration of Nick’s thresholding method.
mance of naive bayes with a hybrid model that integrates Fuzzy [2] research introduces a new convolutional neural net-
C-Means segmentation using two and three cluster configurations work (CNN) model, KR-NET, customized the network model
with nick threshold to KR NET for classifying the flower dataset. for the classification of retinal fundus images for glaucoma
The primary objective of the research is to evaluate each model’s detection. The paper details the application of a 26-layer
performance in handling diverse image datasets, with a focus on
metrics such as accuracy, and precision. CNN that utilizes variable mode decomposition for segmenting
Index Terms—computer vision, naive bayes, hybrid segmenta- the region of interest (ROI) in the fundus images, enhanc-
tion, convolutional neural network, flower dataset. ing the model’s ability to discriminate features crucial for
identifying glaucoma. The authors utilized a 10-fold cross-
I. I NTRODUCTION validation method and a 70:30 split ratio to validate their
In the field of computer vision, the effectiveness of clas- model, achieving notable accuracy rates across various public
sification and segmentation algorithms plays a pivotal role and private datasets.
in enhancing the interpretation and analysis of visual data.
III. M ETHODOLOGY
Traditional classification methods like Naı̈ve Bayes have used
for the long time due to its simplicity and ease of imple- This research has 2 methodologies applied to each approach
mentation. However, with the increasing complexity of image of classification method.
data and the need for higher precision and accuracy, there is a
continuous demand for more new approaches and techniques. A. Naive Bayes Classification
The integration of segmentation techniques can potentially This research employs a hybrid segmentation and classifi-
enhance its effectiveness, especially when applied to large cation methodology for the classification of flowers from the
visual data. in the other side, there is another technique for Oxford 102 Flower Dataset. The image will be segmented us-
classify the image using convolutional neural network (CNN), ing Fuzzy C-means for initial clustering followed by adaptive
CNN are designed to automatically and adaptively learn spatial thresholding to refine the segmentation, Nick adaptive thresh-
hierarchies of features through backpropagation, from low- to olding is applied after the Fuzzy C-means clustering to manage
high-level patterns. This makes them particularly suitable for variable image backgrounds and enhance the differentiation
image classification tasks where the preservation of spatial between each class. After the segmentation, key features such
relationships is key to recognizing visual patterns. as area, perimeter, major and minor axis lengths, aspect ratio,
This research focuses on evaluating the impact of hybrid eccentricity, and solidity are extracted. Then, the feature will
segmentation techniques on a traditional classifier and a convo- be trained for the model. After that, the model will be tried to
lutional neural network (CNN), using the Oxford 102 Flower classify the validation data. The methodology could be seen
Dataset as a test case. at 1

II. L ITERATURE R EVIEW B. KR NET


[1] shows the application of adaptive thresholding tech- While the using of CNN, this research is conduct using
niques combined with Naı̈ve Bayes and K-nearest neighbor a new CNN model known as KR NET. The original idea
classifiers for the classification of red blood cells and sickle of KR NET model is to classifying retinal fundus images
cells. Their research introduces a hybrid segmentation ap- for glaucoma detection, the image need to be applied the
proach, leveraging fuzzy C-means clustering and adaptive grayscaling, hence this research is used to classifying the
thresholding, to enhance the detection accuracy in microscopic flowers, the color information of the flower is important
2), providing a unified metric summarizing the model’s overall
performance (Equation 4).
Input Image
TP + TN
Accuracy = (1)
TP + TN + FP + FN
TP
recall = (2)
TP + FN
TP
precision = (3)
Fuzzy C-Means TP + FP
precision × recall
F 1 − score = 2 × (4)
precision + recall
Hybrid Segmentation
IV. E XPERIMENT AND R ESULT
Adaptive The research employ 2 experimental approach to learn the
Thresholding performance of classification performance. the first approach
is using the hybrid segmentation using k-nearest neighbor and
nick thresholding applied to naive bayes classification, and
the other is using convolutional neural network with KR NET
model.
A. Dataset
Feature In this research the dataset is using the Oxford 102 Flower
Extraction Dataset from kaggle. the dataset has 6552 images for training
and 818 images for validating which separated into 102
classes. Each flower class does not have the same amount of
pictures in the training dataset.
Naive Bayes
B. Experiment 1: Hybrid Segmentation Naive Bayes
Classifier
The first experiment, this research implement the hybrid
segmentation applied to naive bayes classifier. The image from
Fig. 1. Traditional Classification Methodology the dataset segmented using K-nearest neighbor and adaptive
thresholding using nick. we employ 2 configuration in the K-
nearest neighbor using 2 and 3 cluster for the segmentation.
feature for the model. so the preprocess just to convert the As shown in 2a is the original image coming from the Oxford
image into 256x256 pixel, before the image implemented to 102 Flower dataset, then after segmentation using 2 clusters
a 26-layer CNN, which includes six convolution layers, four in K-nearest neighbor, the image is changed to 2b, while in
pooling layers, and one fully connected layer. This network 2c shown the segmented imaged using 3 cluster in the K-
is specifically designed to extract distinctive features from the nearest neighbor. The segmentation shows that using 3 cluster
image and classify them using a softmax layer. the model layer the segmentation locate the objects more than while using 2
of the KR Net is shown in I cluster.
In II shows the result of implementing the hybrid seg-
mentation using K-nearest neighbor with 2 cluster configu-
C. Performance measurement
ration and adaptive thresholding nick that applied to naive
We evaluate classification performance by assessing accu- bayes classifier. in the table shows a wide variance in model
racy, precision, recall, and F1-score based on the classifi- performance across different classes. Classes such as ’bee
cation outcomes. Accuracy, the ratio of correct predictions, balm’ and ’bishop of llandaff’ shows a high recall rates,
is calculated by dividing the sum of true positives and true suggesting effective identification though often at the expense
negatives by the total number of observations (Equation 1). of precision. Conversely, several classes including ’azalea’ and
While accuracy offers an overall performance measure, it may ’balloon flower’ recorded zero values across all performance
not reveal specific issues like a high misclassification rate metrics, indicating areas where the model fails to recognize
for a particular class. Therefore, additional metrics like recall or correctly classify these instances.
and F1-score become essential. Recall measures the number while in III shows the result with 3 cluster configuration.
of correctly predicted positives (Equation 2), while the F1- The result shows the these 2 configuration clusters doesn’t
score combines precision (Equation 3) and recall (Equation affected the performance signifincantly. For instance, class
TABLE I
KR NET ARCHITECTURE LAYER DETAILS FOR 102 CLASSES .

Layer Number of feature maps Stride Activation shape Total learnable parameters
Input 3 - 128 x 128 x 3 -
Convolution 48 1 126 x 126 x 48 1344
Batch normalization 48 - 126 x 126 x 48 96
ReLU 48 - 126 x 126 x 48 -
Max. pooling 48 2 63 x 63 x 48 -
Convolution 48 1 61 x 61 x 48 20784
Batch normalization 48 - 61 x 61 x 48 96
ReLU 48 - 61 x 61 x 48 -
Max. pooling 48 2 30 x 30 x 48 -
Convolution 48 1 4 x 4 x 48 20784
Batch normalization 48 - 4 x 4 x 48 96
ReLU 48 - 4 x 4 x 48 -
Convolution 48 1 2 x 2 x 48 20784
Batch normalization 48 - 2 x 2 x 48 96
ReLU 48 - 2 x 2 x 48 -
Fully connected 102 - 1 x 1 x 102
Softmax 102 - 1 x 1 x 102 -

”bee balm” shows improved recall but decreased precision, it and recall into a single metric, stands at 48.2%. This suggests
suggests the model is identifying more true positives but also that while the model is relatively consistent in identifying
more false positives, which might be a trade-off depending on correct positives.
the application of the algorithm. Classes like ”clematis” that
V. C ONCLUSION
might have improved in both metrics suggest a better fitting
of the model for those classes with an additional cluster. The comparative study on the classification of the Oxford
Among the high performers, ’passion flower’ stands out 102 Flower dataset reveals that the Convolutional Neural
with the highest F1-Score, which can be attributed to its Network (CNN) using the KR NET model significantly out-
distinct features that are easily captured by the model. In performs the hybrid segmentation technique that combines K-
contrast, ’clematis’, with the largest support but zero scores nearest neighbor clustering and nick adaptive thresholding,
across the board, presents a challenge, potentially due to its that applied to a Naive Bayes classifier. Despite the enhance-
visual similarity with other flowers or variability within the ments to the Naive Bayes classifier, its performance did not
class itself. reach that of the CNN, highlighting the substantial gap be-
tween traditional machine learning methods and modern deep
C. Experiment 2: Convolutional neural network using KR learning approaches in handling complex image classification
NET Model tasks.
Then for the second experiment, we doing it using the R EFERENCES
convolutional neural network using KR NET model. adapted to
[1] Chayashree Patgiri and Amrita Ganguly. “Adaptive
[2] the process begin with resize the image and then grayscale
thresholding technique based classification of red blood
the image before input those image into the model. but while
cell and sickle cell using Naı̈ve Bayes Classifier and
this dataset is a flower dataset, the grayscaling technique is
K-nearest neighbor classifier”. In: Biomedical Signal
not applied, the idea is using their color as a feature for the
Processing and Control 68 (2021), p. 102745. ISSN:
model.
1746-8094. DOI: https://doi.org/10.1016/j.bspc.2021.
The IV shows the performance of KR NET classification
102745.
on the Oxford 102 flower dataset. Certain classes like ”colt’s
[2] Kamesh Sonti and Ravindra Dhuli. “A new convolution
foot” and ”fire lily” show fine performance performance with
neural network model “KR-NET” for retinal fundus glau-
F1-scores of 1.0, indicating perfect precision and recall. This
coma classification”. In: Optik 283 (2023), p. 170861.
suggests that the model can distinctly identify features specific
ISSN : 0030-4026. DOI : https://doi.org/10.1016/j.ijleo.
to these classes without any false identifications.
2023.170861.
In contrast, classes like ”anthurium,” ”azalea,” and ”monks-
hood” exhibit zero values across precision, recall, and F1-
score, implying an inability of the model to correctly identify
any instances of these classes. This could be due to a lack of
distinctive features learned by the model for these flowers.
The KR NET model shows a moderate level of accuracy
(53.5%) on a complex task like flower classification, with
a balanced average precision and recall around 55.2% and
52.3% respectively. The F1-score, which combines precision
(a) (b)

(c)
Fig. 2. (a) Original Image. (b) Segmented image using 2 cluster.(c) Segmented image using 3 cluster.
VI. ATTACHMENT

TABLE II: Classification Report for 2 Cluster

Class Precision Recall F1-Score Support


alpine sea holly 0.012349691257718558 0.028044280442804426 0.017148014440433214 1355.0
anthurium 0.022935779816513763 0.007547169811320755 0.011357183418512209 1325.0
artichoke 0.04716981132075472 0.0018037518037518038 0.0034746351633078527 2772.0
azalea 0.0 0.0 0.0 1453.0
ball moss 0.01293103448275862 0.009630818619582664 0.011039558417663294 1869.0
Continued on next page
Class Precision Recall F1-Score Support
balloon flower 0.0 0.0 0.0 595.0
barbeton daisy 0.02185792349726776 0.006756756756756757 0.01032258064516129 2368.0
bearded iris 0.012243113248797552 0.02329450915141431 0.016050444253367727 1202.0
bee balm 0.007246376811594203 0.125 0.0136986301369863 16.0
bird of paradise 0.024396254312469196 0.036450662739322535 0.029229406554472984 2716.0
bishop of llandaff 0.004324843825084094 0.11842105263157894 0.008344923504867872 76.0
black-eyed susan 0.006265664160401002 0.25 0.012224938875305624 40.0
blackberry lily 0.010058675607711651 0.01932367149758454 0.013230429988974642 621.0
blanket flower 0.0 0.0 0.0 1256.0
bolero deep blue 0.0 0.0 0.0 1301.0
bougainvillea 0.017804154302670624 0.056009334889148193 0.0270194202082747 857.0
bromelia 0.011968085106382979 0.0056461731493099125 0.0076726342710997444 1594.0
buttercup 0.0048 0.0316622691292876 0.008336227856894755 379.0
californian poppy 0.005025125628140704 0.03169014084507042 0.008674698795180723 284.0
camellia 0.05660377358490566 0.004834810636583401 0.008908685968819599 1241.0
canna lily 0.0011737089201877935 0.09523809523809523 0.002318840579710145 21.0
canterbury bells 0.0 0.0 0.0 267.0
cape flower 0.010602678571428572 0.021889400921658985 0.014285714285714285 868.0
carnation 0.0029222676797194622 0.015772870662460567 0.004930966469428008 317.0
cautleya spicata 0.012438748586505842 0.04094292803970223 0.019080659150043366 806.0
clematis 0.0 0.0 0.0 3921.0
colt’s foot 0.0058309037900874635 0.011009174311926606 0.007623888182973317 545.0
columbine 0.013100436681222707 0.004720692368214005 0.006940427993059572 1271.0
common dandelion 0.01194921583271098 0.01664932362122789 0.01391304347826087 961.0
corn poppy 0.0011650485436893205 0.2 0.0023166023166023165 15.0
cyclamen 0.03604395604395604 0.03566768160069596 0.03585483165719283 2299.0
daffodil 0.008764940239043825 0.01779935275080906 0.011745862253069941 618.0
desert-rose 0.012012012012012012 0.004524886877828055 0.00657354149548069 884.0
english marigold 0.0068783068783068784 0.03571428571428571 0.011535048802129548 364.0
fire lily 0.006433166547533953 0.008695652173913044 0.007395234182415777 1035.0
foxglove 0.03768115942028986 0.005764966740576497 0.01 2255.0
frangipani 0.01471825063078217 0.023553162853297442 0.018115942028985508 1486.0
fritillary 0.052461139896373056 0.02621359223300971 0.03495899870522227 3090.0
garden phlox 0.0 0.0 0.0 816.0
gaura 0.0 0.0 0.0 527.0
gazania 0.015625 0.013392857142857142 0.014423076923076924 448.0
geranium 0.009783071033602722 0.02077687443541102 0.013302486986697512 1107.0
giant white arum lily 0.01531058617672791 0.027888446215139442 0.019768426998023156 1255.0
globe thistle 0.007224334600760456 0.027181688125894134 0.011414839291078402 699.0
globe-flower 0.0 0.0 0.0 13.0
grape hyacinth 0.005997001499250375 0.024615384615384615 0.009644364074743821 325.0
great masterwort 0.006779661016949152 0.006944444444444444 0.00686106346483705 576.0
hard-leaved pocket orchid 0.009429280397022333 0.02923076923076923 0.01425891181988743 650.0
hibiscus 0.027210884353741496 0.0026702269692923898 0.004863221884498481 2996.0
hippeastrum 0.021235521235521235 0.027066929133858268 0.023799221116399826 2032.0
japanese anemone 0.0 0.0 0.0 456.0
king protea 0.029585798816568046 0.003075030750307503 0.005571030640668524 1626.0
lenten rose 0.0 0.0 0.0 1347.0
lotus lotus 0.01568441064638783 0.024017467248908297 0.01897642323174238 1374.0
love in the mist 0.018364669873196328 0.01991465149359886 0.01910828025477707 2109.0
magnolia 0.0 0.0 0.0 2249.0
mallow 0.0 0.0 0.0 1450.0
Continued on next page
Class Precision Recall F1-Score Support
marigold 0.005219985085756898 0.11864406779661017 0.01 59.0
mexican aster 0.015503875968992248 0.002849002849002849 0.0048134777376654635 702.0
mexican petunia 0.005698005698005698 0.0011869436201780415 0.0019646365422396855 1685.0
monkshood 0.0 0.0 0.0 524.0
moon orchid 0.0004997501249375312 0.0010570824524312897 0.0006786562606040041 946.0
morning glory 0.03372681281618887 0.01554605518849592 0.021282255919127427 2573.0
orange dahlia 0.006224066390041493 0.004137931034482759 0.004971002485501243 725.0
osteospermum 0.006457564575645757 0.009523809523809525 0.007696536558548653 735.0
oxeye daisy 0.0 0.0 0.0 1236.0
passion flower 0.0741042345276873 0.025695326838910067 0.03815913617779641 7083.0
pelargonium 0.0 0.0 0.0 1379.0
peruvian lily 0.011029411764705883 0.0102880658436214 0.0106458481192335 1458.0
petunia 0.0530565167243368 0.012816940651992199 0.02064631956912029 3589.0
pincushion flower 0.009836065573770493 0.014469453376205787 0.011711125569290826 622.0
pink primrose 0.00865265760197775 0.013888888888888888 0.010662604722010662 504.0
pink-yellow dahlia 0.0 0.0 0.0 3077.0
poinsettia 0.0136986301369863 0.003213367609254499 0.0052056220718375845 1556.0
primula 0.00904977375565611 0.013761467889908258 0.01091901728844404 872.0
prince of wales feathers 0.0 0.0 0.0 383.0
purple coneflower 0.0 0.0 0.0 4906.0
red ginger 0.02089009990917348 0.0235655737704918 0.022147327876745306 976.0
rose 0.02241537053979872 0.040731504571903575 0.02891708468574801 1203.0
ruby-lipped cattleya 0.006081337894336754 0.033126293995859216 0.010276172125883108 483.0
siam tulip 0.0020013342228152103 0.012096774193548387 0.0034344590726960505 248.0
silverbush 0.009973149213655544 0.09523809523809523 0.018055555555555554 273.0
snapdragon 0.016181229773462782 0.0029940119760479044 0.005053057099545225 1670.0
spear thistle 0.0 0.0 0.0 327.0
spring crocus 0.0 0.0 0.0 499.0
stemless gentian 0.0 0.0 0.0 1353.0
sunflower 0.025 0.0015426147319706903 0.002905920813657828 2593.0
sweet pea 0.008573178199632579 0.025134649910233394 0.012785388127853882 557.0
sweet william 0.024752475247524754 0.004464285714285714 0.007564296520423601 2240.0
sword lily 0.026022304832713755 0.006286484059272564 0.010126582278481013 2227.0
thorn apple 0.02073050345508391 0.02551640340218712 0.02287581699346405 823.0
tiger lily 0.0 0.0 0.0 132.0
toad lily 0.007590132827324478 0.007905138339920948 0.007744433688286544 506.0
tree mallow 0.0071315372424722665 0.018633540372670808 0.010315186246418338 483.0
tree poppy 0.021052631578947368 0.0020768431983385254 0.003780718336483932 963.0
trumpet creeper 0.0 0.0 0.0 1078.0
wallflower 0.012755102040816327 0.05411255411255411 0.02064409578860446 462.0
water lily 0.02569832402234637 0.028186274509803922 0.026884862653419054 2448.0
watercress 0.012170385395537525 0.061855670103092786 0.020338983050847456 388.0
wild pansy 0.01049317943336831 0.012422360248447204 0.011376564277588168 805.0
windflower 0.0034762456546929316 0.016544117647058824 0.005745292052345995 544.0
yellow iris 0.011042944785276074 0.047872340425531915 0.01794616151545364 564.0
TABLE III: Classification Report for 3 Cluster

Class Precision Recall F1-Score Support


alpine sea holly 0.0 0.0 0.0 1532.0
anthurium 0.03404255319148936 0.0023661638568470865 0.004424778761061947 3381.0
artichoke 0.0379746835443038 0.0020484807101399795 0.003887269193391642 2929.0
azalea 0.0136986301369863 0.001692047377326565 0.0030120481927710845 1773.0
ball moss 0.014193548387096775 0.011305241521068859 0.012585812356979404 1946.0
balloon flower 0.0 0.0 0.0 726.0
barbeton daisy 0.02040816326530612 0.0002874389192296637 0.0005668934240362812 3479.0
bearded iris 0.0 0.0 0.0 4031.0
bee balm 0.0027397260273972603 0.01015228426395939 0.004314994606256742 197.0
bird of paradise 0.01186178442496132 0.008527994067482388 0.00992234685073339 2697.0
bishop of llandaff 0.010109179134654266 0.021367521367521368 0.01372495196266813 1170.0
black-eyed susan 0.0062266500622665 0.0199203187250996 0.009487666034155597 753.0
blackberry lily 0.011126564673157162 0.02564102564102564 0.015518913676042677 936.0
blanket flower 0.0 0.0 0.0 1610.0
bolero deep blue 0.0 0.0 0.0 1573.0
bougainvillea 0.017022961203483768 0.027581783194355357 0.021052631578947368 1559.0
bromelia 0.010396975425330813 0.0040219378427787935 0.005800158186132349 2735.0
buttercup 0.011851851851851851 0.016537467700258397 0.013807982740021575 1935.0
californian poppy 0.005681818181818182 0.01089108910891089 0.00746775288526816 1010.0
camellia 0.0 0.0 0.0 2822.0
canna lily 0.002105263157894737 0.008583690987124463 0.0033812341504649195 233.0
canterbury bells 0.0019011406844106464 0.0045871559633027525 0.002688172043010753 218.0
cape flower 0.008968609865470852 0.00586940572267058 0.007095343680709534 1363.0
carnation 0.0017226528854435831 0.004555808656036446 0.0025 439.0
cautleya spicata 0.011367673179396092 0.03180914512922465 0.016749542004710807 1006.0
clematis 0.027558050523092625 0.022031823745410038 0.024487019612288857 4902.0
colt’s foot 0.0056433408577878106 0.00655307994757536 0.006064281382656155 763.0
columbine 0.007952286282306162 0.0022148394241417496 0.0034647033347769596 1806.0
common dandelion 0.012925969447708578 0.005392156862745098 0.007609823590453131 2040.0
corn poppy 0.010050251256281407 0.004123711340206186 0.005847953216374269 485.0
cyclamen 0.040382889620101704 0.03323485967503693 0.03646185010128292 4062.0
daffodil 0.01570996978851964 0.01808066759388039 0.016812156482379565 1438.0
desert-rose 0.008748317631224764 0.01527614571092832 0.011125374411638854 851.0
english marigold 0.0066181336863004635 0.01152073732718894 0.008406893652795292 868.0
fire lily 0.008481421647819063 0.02001906577693041 0.011914893617021277 1049.0
foxglove 0.019461077844311378 0.005654632448890822 0.008763060330299966 2299.0
frangipani 0.02396898131829397 0.03173121791880541 0.027309236947791166 2143.0
fritillary 0.03371970495258166 0.02249560632688928 0.026987138941598144 2845.0
garden phlox 0.007792207792207792 0.0034482758620689655 0.004780876494023904 870.0
gaura 0.0 0.0 0.0 1480.0
gazania 0.004583333333333333 0.016467065868263474 0.007170795306388526 668.0
geranium 0.022922636103151862 0.017072129748186088 0.019569471624266144 2343.0
giant white arum lily 0.011682242990654205 0.025603511338697878 0.016044006417602567 1367.0
globe thistle 0.0 0.0 0.0 759.0
globe-flower 0.0028169014084507044 0.015037593984962405 0.004744958481613286 133.0
grape hyacinth 0.00532724505327245 0.04154302670623145 0.009443507588532884 337.0
great masterwort 0.005925925925925926 0.0047169811320754715 0.005252790544977019 848.0
hard-leaved pocket orchid 0.007078507078507079 0.01870748299319728 0.01027077497665733 588.0
hibiscus 0.0 0.0 0.0 4012.0
hippeastrum 0.015527950310559006 0.02891459074733096 0.020205160087037613 2248.0
Continued on next page
Class Precision Recall F1-Score Support
japanese anemone 0.0 0.0 0.0 681.0
king protea 0.005109862033725089 0.0045351473922902496 0.004805382027871216 2205.0
lenten rose 0.0 0.0 0.0 1625.0
lotus lotus 0.017977528089887642 0.015497869043006587 0.016645859342488557 2581.0
love in the mist 0.02531645569620253 0.0015238095238095239 0.002874595759971254 2625.0
magnolia 0.0 0.0 0.0 2587.0
mallow 0.027642276422764227 0.013076923076923076 0.017754569190600523 1300.0
marigold 0.0 0.0 0.0 590.0
mexican aster 0.009831460674157303 0.004405286343612335 0.006084311169056932 1589.0
mexican petunia 0.0 0.0 0.0 2261.0
monkshood 0.007575757575757576 0.0014104372355430183 0.0023781212841854932 709.0
moon orchid 0.00034223134839151266 0.0012391573729863693 0.0005363368195226602 807.0
morning glory 0.029209621993127148 0.0049047893825735715 0.008399209486166008 3466.0
orange dahlia 0.0035398230088495575 0.011251758087201125 0.005385392123864019 711.0
osteospermum 0.0055248618784530384 0.006200177147918512 0.005843071786310518 1129.0
oxeye daisy 0.012195121951219513 0.0007830853563038371 0.0014716703458425313 1277.0
passion flower 0.061514997458057954 0.02339068238932921 0.033893557422969185 10346.0
pelargonium 0.010773751224289911 0.005717255717255718 0.007470288624787776 1924.0
peruvian lily 0.011598587997982855 0.012292891501870658 0.011935651271406332 1871.0
petunia 0.053824362606232294 0.003353335686551359 0.0063133410865592295 5666.0
pincushion flower 0.009650180940892641 0.004181913225300575 0.005835156819839533 1913.0
pink primrose 0.0 0.0 0.0 1373.0
pink-yellow dahlia 0.0 0.0 0.0 3981.0
poinsettia 0.0171990171990172 0.0033034450212364322 0.0055423594615993665 2119.0
primula 0.0 0.0 0.0 1955.0
prince of wales feathers 0.002890173410404624 0.0025575447570332483 0.0027137042062415195 391.0
purple coneflower 0.0 0.0 0.0 5286.0
red ginger 0.010050251256281407 0.04887218045112782 0.01667201025969862 532.0
rose 0.030362389813907934 0.037237237237237236 0.03345022929592663 1665.0
ruby-lipped cattleya 0.007168458781362007 0.02618657937806874 0.011255715793176222 611.0
siam tulip 0.0012618296529968455 0.0038461538461538464 0.0019002375296912114 520.0
silverbush 0.009919697685403874 0.02564102564102564 0.01430517711171662 819.0
snapdragon 0.00949667616334283 0.00441696113074205 0.006029544769369913 2264.0
spear thistle 0.0 0.0 0.0 487.0
spring crocus 0.0 0.0 0.0 617.0
stemless gentian 0.015873015873015872 0.0009881422924901185 0.0018604651162790699 1012.0
sunflower 0.0 0.0 0.0 2590.0
sweet pea 0.006653992395437262 0.015069967707212056 0.009231783712495879 929.0
sweet william 0.012587412587412588 0.002922077922077922 0.0047430830039525695 3080.0
sword lily 0.03198976327575176 0.014363688595231256 0.019825535289452814 3481.0
thorn apple 0.025699745547073792 0.03013126491646778 0.02773963196923922 3352.0
tiger lily 0.0020242914979757085 0.005649717514124294 0.0029806259314456036 177.0
toad lily 0.004201680672268907 0.004561003420752566 0.004373974849644614 877.0
tree mallow 0.005681818181818182 0.0006455777921239509 0.0011594202898550724 1549.0
tree poppy 0.0 0.0 0.0 1710.0
trumpet creeper 0.0 0.0 0.0 872.0
wallflower 0.01920860545524395 0.026385224274406333 0.022232103156958647 1895.0
water lily 0.034074074074074076 0.013597398758498374 0.019437988590745827 3383.0
watercress 0.011117612638970159 0.010907003444316877 0.011011301072153 1742.0
wild pansy 0.0156794425087108 0.014106583072100314 0.01485148514851485 1276.0
windflower 0.0 0.0 0.0 1404.0
yellow iris 0.01623119556611243 0.038317757009345796 0.022803114571746386 1070.0
TABLE IV: Classification Report for KR NET

Class Precision Recall F1-Score Support


alpine sea holly 0.3333333333333333 0.5 0.4 4.0
anthurium 0.0 0.0 0.0 12.0
artichoke 1.0 0.4285714285714285 0.6 7.0
azalea 0.0 0.0 0.0 8.0
ball moss 1.0 0.1666666666666666 0.2857142857142857 6.0
balloon flower 0.0 0.0 0.0 4.0
barbeton daisy 0.7142857142857143 0.3125 0.4347826086956521 16.0
bearded iris 0.8 0.5 0.6153846153846154 8.0
bee balm 0.25 0.5 0.3333333333333333 2.0
bird of paradise 0.4444444444444444 0.8 0.5714285714285714 5.0
bishop of llandaff 0.8888888888888888 0.8888888888888888 0.8888888888888888 9.0
black-eyed susan 1.0 0.875 0.9333333333333332 8.0
blackberry lily 0.0 0.0 0.0 6.0
blanket flower 1.0 0.8333333333333334 0.9090909090909092 6.0
bolero deep blue 1.0 0.5 0.6666666666666666 4.0
bougainvillea 1.0 0.0769230769230769 0.1428571428571428 13.0
bromelia 1.0 0.1666666666666666 0.2857142857142857 6.0
buttercup 1.0 0.3333333333333333 0.5 9.0
californian poppy 0.8333333333333334 0.7142857142857143 0.7692307692307693 7.0
camellia 1.0 0.2 0.3333333333333333 10.0
canna lily 0.5 0.5 0.5 2.0
canterbury bells 0.0 0.0 0.0 2.0
cape flower 0.5 0.875 0.6363636363636364 8.0
carnation 0.0 0.0 0.0 2.0
cautleya spicata 0.8571428571428571 1.0 0.9230769230769232 6.0
clematis 0.5555555555555556 0.3846153846153846 0.4545454545454545 13.0
colt’s foot 1.0 1.0 1.0 5.0
columbine 0.0 0.0 0.0 10.0
common dandelion 0.9 0.8181818181818182 0.8571428571428571 11.0
corn poppy 0.0 0.0 0.0 3.0
cyclamen 0.4516129032258064 0.56 0.5 25.0
daffodil 0.25 0.1666666666666666 0.2 6.0
desert-rose 0.2 0.8 0.32 5.0
english marigold 0.4 0.2857142857142857 0.3333333333333333 7.0
fire lily 1.0 1.0 1.0 4.0
foxglove 0.6923076923076923 0.6428571428571429 0.6666666666666666 14.0
frangipani 0.6666666666666666 0.2222222222222222 0.3333333333333333 18.0
fritillary 0.625 0.4166666666666667 0.5 12.0
garden phlox 0.0 0.0 0.0 3.0
gaura 0.6 1.0 0.75 6.0
gazania 0.7142857142857143 1.0 0.8333333333333334 5.0
geranium 0.8125 0.9285714285714286 0.8666666666666667 14.0
giant white arum lily 0.6 0.4285714285714285 0.5 7.0
globe thistle 0.6 0.75 0.6666666666666666 4.0
globe-flower 1.0 0.5 0.6666666666666666 2.0
grape hyacinth 0.3333333333333333 0.5 0.4 2.0
great masterwort 1.0 0.25 0.4 4.0
hard-leaved pocket orchid 0.5 1.0 0.6666666666666666 6.0
hibiscus 0.1818181818181818 0.6153846153846154 0.2807017543859649 13.0
hippeastrum 0.75 0.3333333333333333 0.4615384615384615 9.0
Continued on next page
Class Precision Recall F1-Score Support
japanese anemone 0.5 0.3333333333333333 0.4 3.0
king protea 0.3333333333333333 0.6 0.4285714285714285 5.0
lenten rose 0.0 0.0 0.0 5.0
lotus lotus 0.4545454545454545 0.4545454545454545 0.4545454545454545 11.0
love in the mist 0.6 0.4285714285714285 0.5 7.0
magnolia 0.2307692307692307 1.0 0.375 6.0
mallow 0.5 0.4285714285714285 0.4615384615384615 7.0
marigold 0.3333333333333333 0.3333333333333333 0.3333333333333333 3.0
mexican aster 1.0 0.4285714285714285 0.6 7.0
mexican petunia 0.75 0.6 0.6666666666666666 10.0
monkshood 0.5 0.3333333333333333 0.4 3.0
moon orchid 0.0 0.0 0.0 1.0
morning glory 0.4482758620689655 0.65 0.5306122448979592 20.0
orange dahlia 0.5714285714285714 1.0 0.7272727272727273 4.0
osteospermum 0.2631578947368421 0.8333333333333334 0.4 6.0
oxeye daisy 0.8 1.0 0.8888888888888888 8.0
passion flower 0.6 1.0 0.75 21.0
pelargonium 0.875 0.875 0.875 8.0
peruvian lily 1.0 0.1818181818181818 0.3076923076923077 11.0
petunia 0.6666666666666666 0.4285714285714285 0.5217391304347826 28.0
pincushion flower 0.3 0.375 0.3333333333333333 8.0
pink primrose 1.0 0.375 0.5454545454545454 8.0
pink-yellow dahlia 0.7368421052631579 1.0 0.8484848484848485 14.0
poinsettia 0.5 0.4444444444444444 0.4705882352941176 9.0
primula 0.0 0.0 0.0 9.0
prince of wales feathers 0.3333333333333333 1.0 0.5 1.0
purple coneflower 0.7 0.875 0.7777777777777778 16.0
red ginger 1.0 0.8 0.8888888888888888 5.0
rose 0.5625 0.6 0.5806451612903226 15.0
ruby-lipped cattleya 0.6666666666666666 0.6666666666666666 0.6666666666666666 6.0
siam tulip 0.2 0.3333333333333333 0.25 3.0
silverbush 1.0 0.4 0.5714285714285714 5.0
snapdragon 0.0 0.0 0.0 10.0
spear thistle 0.3333333333333333 1.0 0.5 1.0
spring crocus 0.5 0.5 0.5 2.0
stemless gentian 0.2777777777777778 1.0 0.4347826086956521 5.0
sunflower 0.875 0.7 0.7777777777777778 10.0
sweet pea 0.0 0.0 0.0 6.0
sweet william 0.4 0.4 0.4 10.0
sword lily 0.125 0.0714285714285714 0.0909090909090909 14.0
thorn apple 0.4285714285714285 1.0 0.6 12.0
tiger lily 0.25 1.0 0.4 1.0
toad lily 0.5 0.25 0.3333333333333333 4.0
tree mallow 0.3076923076923077 0.8 0.4444444444444444 5.0
tree poppy 0.8571428571428571 0.8571428571428571 0.8571428571428571 7.0
trumpet creeper 0.6666666666666666 0.4 0.5 5.0
wallflower 0.7857142857142857 0.6111111111111112 0.6875 18.0
water lily 0.4375 0.7368421052631579 0.5490196078431373 19.0
watercress 0.4782608695652174 0.6875 0.5641025641025641 16.0
wild pansy 1.0 0.7 0.8235294117647058 10.0
windflower 1.0 0.4 0.5714285714285714 5.0
yellow iris 0.7142857142857143 0.7142857142857143 0.7142857142857143 7.0

You might also like