Exploring Deep Learning Models for Image Recognition: A Comparative Review

Signal & Image Processing: An International Journal (SIPIJ) Vol.15, No.3, June 2024
DOI: 10.5121/sipij.2024.15301 1
EXPLORING DEEP LEARNING MODELS FOR IMAGE
RECOGNITION: A COMPARATIVE REVIEW
Siddhartha Nuthakki1
, Sai Kalyana pranitha buddiga2
, and Sonika Koganti3
1
Senior Data Scientist, First Object Inc, TX, USA,
2
Independent Researcher, Boston, USA,
3
Software Engineer, Thought Circuits, TX, USA
ABSTRACT
Image recognition, which comes under Artificial Intelligence (AI) is a critical aspect of computer vision,
enabling computers or other computing devices to identify and categorize objects within images. Among
numerous fields of life, food processing is an important area, in which image processing plays a vital role,
both for producers and consumers. This study focuses on the binary classification of strawberries, where
images are sorted into one of two categories. We Utilized a dataset of strawberry images for this study; we
aim to determine the effectiveness of different models in identifying whether an image contains
strawberries. This research has practical applications in fields such as agriculture and quality control. We
compared various popular deep learning models, including MobileNetV2, Convolutional Neural Networks
(CNN), and DenseNet121, for binary classification of strawberry images. The accuracy achieved by
MobileNetV2 is 96.7%, CNN is 99.8%, and DenseNet121 is 93.6%. Through rigorous testing and analysis,
our results demonstrate that CNN outperforms the other models in this task. In the future, the deep
learning models can be evaluated on a richer and larger number of images (datasets) for better/improved
results.
KEYWORDS
Image recognition, Deep Learning, CNN, MobileNetV2, DenseNet121, Binary Classification, Strawberry
Images
1. INTRODUCTION
Computer vision is a sub branch of Artificial Intelligence (AI). In computer vision, image
processing techniques, such as detection and recognition have gained the utmost importance in
different fields of life, such as food industries, and health-related image processing, to name a
few. It includes the automatic categorization of objects within digital images [1, 2]. This
technology has made considerable advancements in recent years, driven by improvements in
Machine Learning (ML) and Deep Learning (DL) algorithms. These advancements have
extended the application spectrum of image detection and recognition, spanning areas such as
agriculture, food processing, autonomous driving, healthcare, and security [3, 4]. The importance
of accurate image detection and recognition systems cannot be ignored, particularly in sectors
where precision is critical [5, 6].
In a Binary classification, a learning technique, the tasks are fundamental yet challenging,
requiring models to distinguish between two classes with good accuracy [7]. Several studies are
available in this connection. For instant, in a study [8], the authors collected a dataset of 11,500
samples from Mendeley and employed various transfer learning models, including VGG161
and
1
https://www.kaggle.com/code/blurredmachine/vggnet-16-architecture-a-complete-guide

2
ResNet502
. Moreover, they presented a hybrid model, VGG16-ResNet that leverages the
advantages of both architectures. By mixing multiple deep learning techniques and transfer
learning strategies, such as VGG16-ResNet50, they developed a robust model capable of
accurately classifying various ranges of dry foods. The approach in this study yielded good
results, achieving a classification accuracy of 99.78% for numerous dry food images, even with
partial training data for the VGG16-ResNet50 model.
Similarly, Y. Kim, et al., 2024 [9], focus on integrated solutions by utilizing computer vision and
deep learning (DL) technologies for ensuring the quality control and optimization of food
processing in complex/hazardous environments. They used the Coffee Bean Classification Model
(CBCM), developed using ML, and demonstrated good performance. It accurately distinguished
coffee beans by navigating around obstacles. The CBCM achieved a maximum validation
accuracy of 98.44% and a minimum validation loss of 5.40% after the 5th
epoch. After tested on a
dataset of 137 samples, the CBCM attained an accuracy of 99.27% and a loss of 2.82%.
In a study on tomatoes [10], introduces a novel approach for tomato quality categorization and
grading, leveraging pre-trained convolutional neural networks (CNNs) for feature extraction and
traditional machine learning algorithms for classification in a hybrid model. Among the proposed
hybrid models, the Convolutional Neural Networks- Support Vector Machine (CNN-SVM)
method outperformed other hybrid approaches, achieving an accuracy of 97.50% in the binary
classification of tomatoes as healthy or rejected and 96.67% in the multiclass classification of
them as ripe, unripe, or rejected when InceptionV3 was used as the feature extractor.
Among delicate fruits, such as tomatoes and Strawberries are a significant crop in different
regions of the world and are considered as most fragile. From 2017 to 2019, Florida's strawberry-
planted area was second only to California, accounting for 20% of the total strawberry-planted
area in the United States (US). Strawberry yield prediction is crucial for farmers, as it enables
improved management of yield logistics, appropriate harvesting decisions, suitable labor hiring,
and profit estimation [11]. Short-term strawberry yields can be accurately predicted by
identifying and estimating the number of strawberries at different maturity stages [12].
To precisely classify the strawberry fruits based on advanced deep learning models, we need to
compare the results generated by numerous deep learning models. This may help the partitioners
to identify the raw and mature strawberry fruits accurately. We employed and compared several
state-of-the-art deep learning models, including MobileNetV2, Convolutional Neural Networks
(CNN), and DenseNet121, for the binary classification of strawberry images. MobileNetV2 is a
well-known model for its efficiency and effectiveness in mobile and embedded vision
applications [13]. CNNs have been the backbone of many image recognition breakthroughs due
to their robust architecture for learning spatial hierarchies in images [14]. DenseNet121, with its
tightly connected convolutional layers, promises efficient parameter usage and mitigation of the
vanishing-gradient problem [15]. These models are chosen for their diverse architectures and
proven performance in various image classification tasks.
This study focuses on a specific aspect of image recognition known as binary classification,
which simplifies the categorization task by sorting images into one of two distinct groups. This
study utilizes a dataset of strawberry images to explore the efficacy of various deep-learning
models in correctly identifying whether an image contains strawberries. The practical
applications of this research are significant, particularly in food processing, and agriculture,
where automated systems for fruit detection can enhance efficiency and accuracy in processes
like collecting and quality control [16]. The ability to accurately classify images as containing
2
https://keras.io/api/applications/resnet/

3
strawberries or not can reduce labor costs and improve the consistency of agricultural outputs.
Through rigorous testing and analysis, our study reveals that the CNN model significantly
outperforms MobileNetV2 and DenseNet121 in the task of binary classification of strawberry
images. The accuracy achieved by MobileNetV2 was 96.7%, DenseNet121 scored 93.6%, and
CNN reached an impressive 99.8%. These results underscore the superiority of CNNs in this
specific context, highlighting their potential for deployment in real-world agricultural
applications [17]. This paper aims to provide a detailed comparative analysis of these models,
contributing valuable insights for practitioners seeking to implement deep learning solutions for
image recognition tasks in various industries.
The rest of the paper is structured as follows: The Literature Review discusses a detailed yet
comprehensive relevant literature. The Methods and Materials discusses the procedure carried out
for this study. Results and Analysis depicted the results. The Conclusion part concludes the
paper. The References are enlisted/enumerated at the end.
2. LITERATURE REVIEW
The precise detection and classification of various fruit species, and food in general, has become
increasingly important [18, 19]. This area holds significant relevance not only in academic
research but also in industrial applications. Many practical applications can be developed based
on such classification systems. One prominent application involves its use in supermarkets to
assist cashiers. Cashiers must accurately identify both the species and variety of fruits purchased
by customers to determine the correct pricing. The pricing information must be stored in a query
table. A classification-based application can automatically identify the fruit species purchased by
the customer and match it with the correct price. This issue has been effectively addressed
through barcode reader systems for packaged products. However, this solution does not apply to
fruits and vegetables because customers typically select each piece of fruit individually when
shopping in supermarkets.
The fruit classification system can also be implemented as a smartphone application. Numerous
commercial and free applications (apps in short) are available in the market [20, 21]. These
applications may assist individuals, particularly those with health concerns, in determining
whether a specific fruit or vegetable meets their dietary needs. The application will identify the
fruit species and present its corresponding nutritional information, essential details, and
recommendations.
Unfortunately, fruit classification based on computer vision presents significant challenges due
to:
 Resemblances in shape, color, and texture.
 Considerable differences within a single fruit category.
Numerous systems discussed in the literature aim to automate fruit inspection for defects, identify
maturity stages, and recognize categories. For instance, a survey on fruit ripeness is discussed in
[22]. Similarly, S. Marimuthu et al, [23] introduces a fuzzy model for classifying pineapple
ripeness levels. This model extracts features such as peak hue and normalized brown area from
the hue channel. Parameter tuning using Particle Swarm Optimization (PSO) yielded an accuracy
of 93.11% on the MUSA database [24], comprising 3108 pineapple images at various ripening
stages. Additionally, the same optimization technique was employed to parallelize computing
resources for faster responses in a study [25].

4
Y. Zhang et al., [26] explored multi-class Kernel SVMs with appearance descriptors for fruit
classification, achieving an accuracy of 88.2% on a dataset of 18 classes with 1653 color images
using a combination of color, texture, and shape descriptors. Similarly, a hybrid classification
approach proposed in [27], feedforward neural network classifiers trained via fitness-scaled
chaotic artificial bee colony optimization achieved an accuracy of 89.1% on the same dataset,
incorporating color histogram, texture, and shape descriptors.
Table 1: Relevant studies
Title Year Key Contributions Methods/Techniques Used
[12] 2019
Introduced a novel deep learning-
based method for classifying
strawberry diseases with enhanced
accuracy, leveraging cloud services
for processing and analysis.
Improved residual networks,
cloud services, data
augmentation, image
preprocessing, transfer
learning
[28] 2021
Developed a system for automatic
disease recognition in crops using
CNNs, achieving high accuracy
across multiple crop species and
diseases.
Convolutional Neural
Networks (CNNs), transfer
learning, image
augmentation, training on
large datasets
[29] 2022
Proposed a self-supervised multi-
network fusion model for detecting
strawberry diseases, improving
classification performance through
network fusion techniques.
Self-supervised learning,
multi-network fusion,
transfer learning, data
augmentation, deep learning
models including CNNs and
residual networks, hybrid
model fusion
3. METHODS AND MATERIAL
This study focuses on a specific type of image recognition called binary classification. This
means we want to sort images into one of two groups. For our experiment, we use a dataset of
strawberry3
images to see how well different models can tell if an image contains strawberries or
not. Comparing several popular models, we aim to find out which ones work best for identifying
strawberries. This can be very useful in areas like farming and quality control. Through careful
testing and analysis of results, we will highlight which models are most effective for this task.
We used MobileNetV2 for the Binary Classification of Strawberry Images, Convolutional Neural
Network (CNN) for the Binary Classification of Strawberry Images, and DenseNet121 for the
Binary Classification of Strawberry Images. All these algorithms are compared. Results show
that CNN outperforms other algorithms. Figure 1 show the workflow of machine and deep
learning models.
3 https://www.kaggle.com/datasets/abdulbasit31/strawberry-dataset/data

5
Figure 1: Workflow of machine and deep learning models [30]
4. RESULTS AND ANALYSIS
In this section, the results are analyzed. The three algorithms are evaluated based on a single
dataset. The following sub-section demonstrated the results as:
5. MOBILENETV2 FOR BINARY CLASSIFICATION OF STRAWBERRY IMAGES
The MobileNetV2 model is initialized with pre-trained weights from the ImageNet dataset and
adapted for binary classification by removing the top layers responsible for classifying ImageNet
categories.
Introduction to MobileNetV2: MobileNetV2 is a lightweight convolutional neural network
architecture designed for efficient image classification tasks, particularly suited for deployment
on mobile and embedded devices. Its architecture consists of depth-wise separable convolutions
and inverted residual blocks, allowing for effective feature extraction while minimizing
computational resources. In this study, we employ MobileNetV2 as a feature extractor for the
binary classification of strawberry images.
Fig 2: The schematic diagram of the proposed solution.

6
A custom classification layer is added on top of the base MobileNetV2 architecture to classify
images as either containing pickable or unpickable strawberries. During training, the base
MobileNetV2 layers are frozen to retain learned features, while only the classification layer is
trained on the strawberry dataset. This transfer learning approach enables the model to quickly
adapt to the new task with relatively few training examples.
Results and Performance Evaluation: Training the MobileNetV2 model on a dataset of strawberry
images results in impressive performance metrics. After 20 epochs of training with data
augmentation techniques applied to increase robustness, the model achieves a training accuracy
of approximately 96.7% and a training loss of 0.193. The validation accuracy is around 80.6%,
with a validation loss of 0.400. Precision, recall, and F1-score metrics are calculated for both
classes, showing the model's ability to effectively distinguish between pickable and unpickable
strawberries. The confusion matrix visualizes the model's predictions compared to the true labels,
revealing few misclassifications and overall strong performance.
6. CONVOLUTIONAL NEURAL NETWORK (CNN) FOR BINARY
CLASSIFICATION OF STRAWBERRY IMAGES
Introduction: In this study, a custom Convolutional Neural Network (CNN) is employed for
binary classification of strawberry images. CNNs are particularly well-suited for image
recognition tasks due to their ability to automatically learn spatial hierarchies of features through
convolutional layers. The objective is to classify images into two categories: "Pickable" (images
containing pickable strawberries) and "UnPickable" (images without pickable strawberries). The
CNN model is designed with several layers to effectively capture and learn features from the
strawberry images:
Table 2: CNN model structure
Input Layer: A Conv2D layer with 32 filters of size 3x3 and ReLU activation
function processes the input images of size 224x224x3.
Convolutional Blocks: The model includes multiple convolutional blocks with
increasing filter sizes (64 and 128) and ReLU activation
functions.
Each block consists of a Conv2D layer followed by Batch
Normalization, MaxPooling2D for down-sampling, and Dropout
to prevent overfitting.
Flatten Layer: The Flatten layer converts the 2D feature maps into a 1D feature
vector.
Fully Connected Layers: A Dense layer with 512 units and ReLU activation is added,
followed by a Dropout layer.
Output Layer: A Dense layer with a single unit and sigmoid activation function
provides the binary classification output.
Training Strategy: The model is trained using the Adam optimizer with a learning
rate of 0.0001 and binary cross-entropy loss.
The data augmentation techniques such as rotation, shifting, shearing, zooming, and horizontal
flipping are applied to the training set to improve generalization. The dataset is split into 80%
training and 20% validation subsets using Image Data Generator. The results show that Training
Loss is about 0.062, where training accuracy is 99.8%. The validation loss was around 1.196
whereas the validation accuracy is 96.1%.

7. DENSENET121 FOR BINARY CLASSIFICATION OF STRAWBERRY IMAGES
Introduction: DenseNet121 is a state-of-the-art convolutional neural network (CNN) architecture
designed to facilitate efficient feature reuse through dense connectivity between layers. This
model is particularly effective for image classification tasks due to its ability to capture intricate
details and hierarchical features. The DenseNet121 is employed for the binary classification of
strawberry images, distinguishing between "Pickable" and "UnPickable" categories.
Table 3:DenseNet121 model structure
Model
Structure
The DenseNet121 model is adapted for this binary classification
task by leveraging transfer learning
Base Model: The DenseNet121 model is initialized with pre-trained weights
from the ImageNet dataset.
The top layers of DenseNet121, which are responsible for
classifying ImageNet categories, are excluded.
Custom
Layers:
A GlobalAveragePooling2D layer is added to reduce the
dimensionality of the feature maps output by DenseNet121.
A Dense layer with a single unit and sigmoid activation
function is used to perform binary classification.
Freezing Base
Layers:
The base layers of DenseNet121 are frozen during training to
preserve the learned features from ImageNet, which helps in
achieving faster convergence and better generalization with
limited training data.
Data
Augmentation:
The training data is augmented using techniques like rotation,
width and height shifts, shearing, zooming, and horizontal
flipping to enhance the model's robustness.
The dataset is split into 80% training and 20% validation subsets using Image Data Generator.
The model is compiled with the Adam optimizer (learning rate: 0.0001) and binary cross-entropy
loss. The model is trained for 20 epochs. We achieved a training accuracy: of 93.6% at the loss of
training Loss: 0.369. Similarly, validation accuracy was 83.5% with a validation loss of 0.423.
Fig 2. Comparative analysis of models based on accuracy

8
Figure 2 shows a comparative analysis of all three models evaluated in this study. Results show
that CNN outperforms the rest of the models. Results may be different on other datasets and
approaches.
8. CHALLENGES
A few challenges are still necessary to explain like the limited size datasets for such
classification. The deep learning algorithms are data-hungry for accurate dictation and
classification [31]. Moreover, the classification of fruits such as strawberries, and apples is easy
to detect due to their colors and similarities in features; however, more entropy in data may lead
to inaccurate results [32]. Furthermore, the model may perform well in controlled environments,
such as theoretically, however, its transferability to real-world scenarios with varying lighting
conditions, and diverse backgrounds needs to be validated. Ensuring the model's robustness
across different real-world conditions is essential for practical deployment.
9. FUTURE WORK
The evaluation of the models on a richer and larger dataset that includes more images captured
under various conditions, such as different lighting, angles, and backgrounds may improve
generalizability and robustness. Extending the research to multi-class classification, instead of
binary classification, where the model not only identifies the presence of strawberries but also
distinguishes between different types of fruits or diverse quality grades of strawberries.
Moreover, utilizing transfer learning to adapt models pre-trained on large datasets to the task of
strawberry classification, could improve performance. Furthermore, using ensemble methods to
combine the strengths of multiple models, such as MobileNetV2, CNN, and DenseNet121, to
achieve improved performance and consistency. The integration of image data with other types of
sensor data (e.g., high pixel camera, temperature, humidity may improve the accuracy and
robustness of the classification models. Tailoring these models and solutions for specific
agricultural applications, such as automated sorting and grading systems to address particular
needs and challenges in the fields are some possible future works.
10. CONCLUSION
In conclusion, image recognition is pivotal in computer vision, allowing computers to identify
and categorize objects within images. This study focused on the binary classification of
strawberry images, assessing the effectiveness of various models in determining the presence of
strawberries. The research holds significant practical applications in agriculture and quality
control. For this classification task, we evaluated several prominent models, including
MobileNetV2, Convolutional Neural Networks (CNN), and DenseNet121. The results revealed
that MobileNetV2 achieved an accuracy of 96.7%, DenseNet121 attained 93.6%, and CNN
excelled with an impressive accuracy of 99.8%. Through rigorous testing and analysis, it is
evident that CNN outperforms the other models, establishing its superiority in the binary
classification of strawberry images. The model can be used for many other similar fruits and
vegetables, such as tomatoes, apples, oranges, to name a few. Moreover, this type of model can
be used for quality control and sorting of various fruits/vegetables in food processing plants. This
may help guarantee reliable quality products and reduce waste.
CONFLICTS OF INTEREST
This section is compulsory. A competing interest exists when professional judgment concerning
the validity of research is influenced by a secondary interest, such as financial gain. We require

that our authors reveal any possible conflict of interest in their submitted manuscripts. If there is
no conflict of interest, authors should state that “The author(s) declare(s) that there is no conflict
of interest regarding the publication of this paper.”
FUNDING STATEMENT
Authors should state how the research and publication of their article was funded, by naming
financially supporting bodies followed by any associated grant numbers in square brackets.
ACKNOWLEDGMENTS
An Acknowledgements section is optional and may recognize those individuals who provided
help during the research and preparation of the manuscript. Other references to the title/authors
can also appear here, such as “Author 1 and Author 2 contributed equally to this work.
REFERENCES
[1] V. Alves, J. M. dos Santos, E. Pinto, I. M. Ferreira, V. A. Lima, and M. L. Felsner, "Digital image
processing combined with machine learning: A new strategy for brown sugar classification,"
Microchemical Journal, vol. 196, p. 109604, 2024.
[2] Nuthakki, S., Kumar, S., Kulkarni, C. S., & Nuthakki, Y. (2022). “Role of AI Enabled Smart Meters
to Enhance Customer Satisfaction”. International Journal of Computer Science and Mobile
Computing, Vol.11 Issue.12, December- 2022, pg. 99-107, doi:
https://doi.org/10.47760/ijcsmc.2022.v11i12.010
[3] A. Krizhevsky, I. Sutskever, and G. E. Hinton, "ImageNet classification with deep convolutional
neural networks," Communications of the ACM, vol. 60, pp. 84-90, 2017.
[4] Kathiriya, S., Nuthakki, S., Mulukuntla, S., & Charllo, B. V. “AI and The Future of Medicine:
Pioneering Drug Discovery with Language Models”, International Journal of Science and
Research, vol. 12, no. 3, pp. 1824-1829, Mar. 2023, doi:
https://dx.doi.org/10.21275/SR24304173757
[5] M. Reddy, A. Bodepudi, M. Mandapuram, and S. S. Gutlapalli, "Face detection and recognition
techniques through the Cloud Network: An Exploratory Study," ABC Journal of Advanced
Research, vol. 9, pp. 103-114, 2020.
[6] Nuthakki, S., Uttiramerur, A. D., Nuthakki, Y., & Munjala, M. B. Navigating the Medical
Landscape: A Review of Chatbots for Biomarker Extraction from Diverse Medical Report,”
International Journal For Multidisciplinary Research, vol. 6, pp. 1-16, 2024, doi:
https://doi.org/10.36948/ijfmr.2024.v06i01.13154.
[7] S. Panopoulos, "Video binary classification using deep learning techniques," Πανεπιστήμιο
Πειραιώς, 2024.
[8] S. N. Nobel, M. A. H. Wadud, A. Rahman, D. Kundu, A. A. Aishi, S. Sazzad, et al., "Categorization
of Dehydrated Food through Hybrid Deep Transfer Learning Techniques," Statistics, Optimization
& Information Computing, 2024.
[9] Y. Kim, J. Lee, and S. Kim, "Study of active food processing technology using computer vision and
AI in coffee roasting," Food Science and Biotechnology, pp. 1-8, 2024.
[10] H. S. Mputu, A. Abdel-Mawgood, A. Shimada, and M. S. Sayed, "Tomato Quality Classification
based on Transfer Learning Feature Extraction and Machine Learning Algorithm Classifiers," IEEE
Access, 2024.
[11] T. Horie, M. Yajima, and H. Nakagawa, "Yield forecasting," Agricultural systems, vol. 40, pp. 211-
236, 1992.
[12] Y. Chen, W. S. Lee, H. Gan, N. Peres, C. Fraisse, Y. Zhang, et al., "Strawberry yield prediction
based on a deep neural network using high-resolution aerial orthoimages," Remote Sensing, vol. 11,
p. 1584, 2019.
[13] M. Sandler, A. Howard, M. Zhu, A. Zhmoginov, and L.-C. Chen, "Mobilenetv2: Inverted residuals
and linear bottlenecks," in Proceedings of the IEEE conference on computer vision and pattern
recognition, 2018, pp. 4510-4520.

10
[14] Suyash Bhogawar, C. A., Nuthakki, S., Venugopal, S. M., & Mullankandy, S. "The Ethical and
Social Implications of Using AI in Healthcare-A Literature Review", International Journal of
Science and Research, Vol.12, no. 11, pg. 1472-1477, Nov. 2023, doi: https:// :
https://dx.doi.org/10.21275/SR231116135559
[15] G. Huang, Z. Liu, L. Van Der Maaten, and K. Q. Weinberger, "Densely connected convolutional
networks," in Proceedings of the IEEE conference on computer vision and pattern recognition,
2017, pp. 4700-4708.
[16] S. Bargoti and J. Underwood, "Deep fruit detection in orchards," in 2017 IEEE international
conference on robotics and automation (ICRA), 2017, pp. 3626-3633.
[17] Nuthakki, S., Bhogawar, S., Venugopal, S. M., & Mullankandy, S. "Conversational AI AND LLM'S
Current And Future Impacts In Improving And Scaling Health Services," International Journal of
Computer Engineering and Technology (IJCET), Vol. 14, no. 3, pp.149-155, Dec. 2023,
https://iaeme.com/Home/issue/IJCET?Volume=14&Issue=3
[18] Y. Zhang, L. Deng, H. Zhu, W. Wang, Z. Ren, Q. Zhou, et al., "Deep learning in food category
recognition," Information Fusion, vol. 98, p. 101859, 2023.
[19] Nuthakki, S., Neela, S., Gichoya, J. W., & Purkayastha, S. (2019). Natural language processing of
MIMIC-III clinical notes for identifying diagnosis and procedures with neural networks. arXiv
preprint arXiv:1912.12397.
[20] A. Appadoo, Y. Gopaul, and S. Pudaruth, "FruVegy: An Android App for the Automatic
Identification of Fruits and Vegetables using Computer Vision and Machine Learning,"
International Journal of Computing and Digital Systems, vol. 13, pp. 169-178, 2023.
[21] Singh, D., Nuthakki, S., Naik, A., Mullankandy, S., Singh, P. K., & Nuthakki, Y. (2022).
“Revolutionizing Remote Health: The Integral Role of Digital Health and Data Science in Modern
Healthcare Delivery”, Cognizance Journal of Multidisciplinary Studies, Vol.2, Issue.3, March 2022,
pg. 20-30, doi: https://10.47760/cognizance.2022.v02i03.002
[22] M. Rizzo, M. Marcuzzo, A. Zangari, A. Gasparetto, and A. Albarelli, "Fruit ripeness classification:
A survey," Artificial Intelligence in Agriculture, vol. 7, pp. 44-57, 2023.
[23] S. Marimuthu and S. M. M. Roomi, "Particle swarm optimized fuzzy model for the classification of
banana ripeness," IEEE Sensors Journal, vol. 17, pp. 4903-4915, 2017.
[24] J. W. Gichoya, S. Nuthakki, P. G. Maity, and S. Purkayastha, “Phronesis of AI in radiology:
Superhuman meets natural stupidity,” arXiv.org, Mar. 27, 2018. https://arxiv.org/abs/1803.11244.
[25] M. S. Hossain, M. Moniruzzaman, G. Muhammad, A. Ghoneim, and A. Alamri, "Big data-driven
service composition using parallel clustered particle swarm optimization in mobile environment,"
IEEE Transactions on Services Computing, vol. 9, pp. 806-817, 2016.
[26] Y. Zhang and L. Wu, "Classification of fruits using computer vision and a multiclass support vector
machine," sensors, vol. 12, pp. 12489-12505, 2012.
[27] Y. Zhang, S. Wang, G. Ji, and P. Phillips, "Fruit classification using computer vision and
feedforward neural network," Journal of Food Engineering, vol. 143, pp. 167-177, 2014.
[28] A. Abade, P. A. Ferreira, and F. de Barros Vidal, "Plant diseases recognition on images using
convolutional neural networks: A systematic review," Computers and Electronics in Agriculture,
vol. 185, p. 106125, 2021.
[29] D Singh, S Bhogawar, S Nuthakki, N Ranganathan, "Enhancing Patient-Centered Care in Oncology
through Telehealth: Advanced Data Analytics and Personalized Strategies in Breast Cancer
Treatment", International Journal of Science and Research (IJSR), Volume 10 Issue 9, September
2021, pp. 1707-1715, https://www.ijsr.net/getabstract.php?paperid=SR24108012724
[30] M. P. Akhter, Z. Jiangbin, I. R. Naqvi, M. Abdelmajeed, A. Mehmood, and M. T. Sadiq,
"Document-level text classification using single-layer multisize filters convolutional neural
network," IEEE Access, vol. 8, pp. 42689-42707, 2020.
[31] A. Kamilaris and F. X. Prenafeta-Boldú, "Deep learning in agriculture: A survey," Computers and
electronics in agriculture, vol. 147, pp. 70-90, 2018.
[32] V. Singh and A. K. Misra, "Detection of plant leaf diseases using image segmentation and soft
computing techniques," Information processing in Agriculture, vol. 4, pp. 41-49, 2017.

Exploring Deep Learning Models for Image Recognition: A Comparative Review

More Related Content

Exploring Deep Learning Models for Image Recognition: A Comparative Review