Identifying New Species of Dogs Using Machine Learning Model
Smita Thube
1
, Sonam Singh
2
, Poonam Sadafal
3
, Shweta Lilhare
4
, Pooja Mohbansi
5
,
Vishal Borate
3 a
and Yogesh Mali
6 b
1
Department of Computer Engineering, Nutan Maharashtra institute of Engineering & Technology, Pune, India
2
Department of Artificial Intelligence and Data Science, Dr. D.Y Patil Institute of Technology, Pimpri, Pune, India
3
Department of Computer Engineering, Dr. D. Y. Patil College of Engineering and Innovation, Talegaon, Pune, India
4
Department of AIML, G.H. Raisoni College of Engineering, Pune, India
5
Department of Computer Engineering, Ajeenkya DY Patil School of Engineering, Lohegaon, Pune, India
6
School of Engineering, Ajeenkya DY Patil University Lohegaon, Pune, India
Keywords: Dog Breed, CNN, SVM, InceptionV3, Xception, VGG, Deep Learning, Machine Learning.
Abstract: This paper addresses the challenging problem of breed identification in dogs, whose applications will be very
important in disease prevention, genetic research, and personal pet care. We here present an advanced system
that identifies dog breeds, using the capabilities of particular CNNs such as InceptionV3, VGG16, Xception,
and ResNet for efficient feature extraction. This classification is then refined by a Support Vector Machine
algorithm to enhance accuracy. The system is trained on the Stanford Dogs Dataset, a rich collection of diverse
dog breed images. The dataset enhances the model's ability to extract meaningful features and classify
accurately a wide variety of dog breeds. By iteratively training the model, it learns subtle breed-specific
patterns in the images and achieves high classification accuracy at 96.3%. This research not only pushes
forward the capabilities of breed identification systems but also offers a flexible approach that can be applied
to various practical scenarios where precise breed recognition is critical. With accuracy and adaptability, our
system is promising for more extensive applications in biology, veterinary science, and personalized pet
management, which would be helpful for insights in the care and research of canines.
1 INTRODUCTION
Dogs play an essential role in human life today. They
can be companions, workers, or therapy animals. The
over 120 distinct breeds with their set of physical
characteristics and behaviours require identification
of the exact breed for several purposes: population
management, veterinary care, animal shelters, and
canine research (More, Jadhav, et al., 2024). Breed
identification, when accurate, is important for several
reasons. New development in CNN has brought the
potential of deep learning in image classification,
such as dog breed categorization (Mali, Pawar, et al.,
2023). This complex nature makes it a perfect
machine learning problem with deep learning
approaches, which learn the complex visual patterns
in the images. Despite these modern advances,
traditional solutions, which work on human
evaluation, are faulty and biased in many cases.
a
https://orcid.org/0009-0009-7585-6667
b
https://orcid.org/0009-0004-0582-9595
Therefore, contemporary approaches are of
importance (Mali, Mohanpurkar, et al., 2015).
CNNs, being a type of deep learning model that
can classify images, have been considered to be an
effective alternative for breed identification since it
provides accuracy and eliminates subjective judgment
(Nalawade, Pattnaik, et al., 2024). The major
challenge is developing a reliable, automated system
that identifies dog breeds with the large variability in
breed characteristics (Patil, Zurange, et al., 2024).
Most methods of breed identification that depend on
manual assessment or predefined rules fail to account
for the scope and variety of dog breeds. However,
accurate breed identification has far- reaching effects-
from optimal management of diseases to advancement
in genetic studies-all the way to the welfare of animals,
the scrutiny of law (Modi, Modi, Shiqi, 2024).
Described previously, advancements in this area
will lead to increased efficiency in animal shelters,
personalized veterinary care, and support for
responsible breeding practices.
Thube, S., Singh, S., Sadafal, P., Lilhare, S., Mohbansi, P., Borate, V. and Mali, Y.
Identifying New Species of Dogs Using Machine Learning Model.
DOI: 10.5220/0013589200004664
Paper published under CC license (CC BY-NC-ND 4.0)
In Proceedings of the 3rd International Conference on Futuristic Technology (INCOFT 2025) - Volume 2, pages 195-202
ISBN: 978-989-758-763-4
Proceedings Copyright © 2025 by SCITEPRESS – Science and Technology Publications, Lda.
195

2 LITERATURE REVIEW
A technique has been introduced for identifying dog
breeds by utilizing Convolutional Neural Networks
(CNNs). This approach employs the TensorFlow and
Keras frameworks to develop and train a model using
a dataset of dog images (Mehta, Chougule, et al.,
2024). This approach achieved a test accuracy of
88.4%, providing a strong introduction to CNN-based
dog breed classification by offering clear insights into
both the methodology and results (Shimpi, Balinge, et
al., 2024).
Another deep learning approach used transfer
learning to enhance performance by fine-tuning pre-
trained CNN models (Ingale, Wankar, et al., 2024).
By preserving these models, the system attained an
accuracy of 89.92% on a dataset containing 133 dog
breeds. This result highlights how transfer learning
enhances CNN performance compared to building a
network from the ground up (Mulani, Nandgaonkar,
et al., 2024).
In addition, CNN and transfer learning were
employed to develop an Android application capable
of identifying a dog's breed from a photo (Sonawane,
Mulani, et al., 2024). Users can upload or take a
picture of a dog, which is then processed to extract
necessary features for classification, achieving 94%
accuracy (Mandale, Modi, et al., 2024). The authors
demonstrate that models like ResNet50 are
particularly effective for dog breed identification due
to their ability to capture intricate details in images
(Sengupta, Nalawade, et al., 2024).
Further research has shown that deep learning
approaches, such as InceptionV3, are very effective
in identifying dog breeds (More, Khane, et al., 2024).
These models are excellent at learning intricate
features from images and achieving high
classification accuracy, with InceptionV3 being
particularly noted for its ability to handle diverse and
complex image details (Wanaskar, Dangore, et al.,
2024).
Another approach is to classify the breeds by
studying the size and position of some parts of the dog
in the image (More, Ramishte, et al., 2024). This
approach employed CNNs for feature extraction and
classification of breeds, obtaining 90.6% accuracy on
the test set (Palkar, Jain, et al., 2024).
Furthermore, the ability to classify breeds using
deep learning to create mobile applications has been
quite useful for dog lovers. Among CNN
architectures, such as VGGNet, ResNet50, and
InceptionV3, the achievement of perfect results is
observed (Dangore, Modi, et al., 2024). Xception is
extraordinary in terms of its ability to learn complex
patterns, making it a good choice for this task as well
(Dangore, Bhaturkar, et al., 2024).
Overall, these methods demonstrate how CNNs, in
combination with approaches such as transfer learning
and mobile integration, are revolutionizing dog breed
identification by providing better accuracy, efficiency,
and accessibility for a wide range of users (More,
Shinde, et al., 2024).
3 METHDOLOGY
The flowchart illustrates the steps involved in building
a dog breed classification model.
Figure 1: Proposed system Block diagram.
3.1 Data Collection and Preparation
A comprehensive dataset with labeled dog images is
gathered, such as the Stanford Dogs Dataset (Vaidya,
Dangore, et al., 2024).
The dataset is divided into:
3.1.1 Training Set
Majority of images used to train the model.
3.1.2 Testing Set
A smaller subset reserved for evaluation.
A CSV file links image IDs to their breed labels
for streamlined data management.
3.2 Data Pre-Processing
Images are resized to a uniform size to align with the
model's input requirements (Sawardekar, Mulla, et al.,
INCOFT 2025 - International Conference on Futuristic Technology
196

2025). Pixel values are normalized to a consistent
range, enhancing model efficiency during training.
Data Augmentation techniques such as rotation,
flipping, zooming, and shifting are applied to
diversify the dataset and improve generalization
(Modi, Mali, et al., 2023).
3.3 Model Training Using Pre-Trained
Networks
Transfer learning leverages pre-trained CNN
architectures like InceptionV3, VGG16, and ResNet
to extract features. Support Vector Machine (SVM) is
used for classification, effectively separating breeds
in high-dimensional feature space (Bhongade,
Dargad, et al., 2024).
3.4 Training Process
3.4.1 Hyperparameter Tuning
Adjustments to learning rate, optimizer, and batch
size optimize training performance.
Validatio
n is performed on a subset of training data to prevent
overfitting. Fine-tuning of CNN layers ensures the
model adapts specifically to dog breed identification
(Mali, Yogesh., et al., 2023).
3.5 Model Evaluation and Testing
The model is tested on unseen images from the testing
set. Performance is measured using:
3.5.1 Accuracy
Percentage of correct predictions.
3.5.2 Log Loss:
Penalizes incorrect predictions more heavily.
3.5.3 Confusion Matrix
Highlights patterns of misclassification among
breeds.
3.6 Model Deployment
Optimization ensures the model is lightweight and
efficient for deployment. Deployment options
include:
3.6.1 Local Deployment
Suitable for offline use.
3.6.2 Cloud Deployment
Accessible via the internet.
A user-friendly interface is created for uploading
images and receiving instant breed predictions. This
structured approach ensures a robust, accurate, and
deployable dog breed classification model (Inamdar,
Faizan, et al., 2024).
4 MODEL ARCHITECTURE
Figure 2 illustrates the structure of the deep learning
model designed for dog breed identification (Jagdale,
Sudarshan, et al., 2020). The architecture consists of
several convolutional layers (Conv2D), batch
normalization layers, and activation functions
(LeakyReLU), pooling layers (MaxPooling2D), and
dropout layers to minimize overfitting (Modi, Mali, et
al., 2023). The model gradually reduces the spatial
dimensions while increasing the depth, allowing it to
learn complex features at each layer (Mali, Sharma, et
al., 2023). Each layer plays a specific role: the
convolutional layers extract features, batch
normalization stabilizes learning, and dropout layers
help make the model more robust (Modi, 2024). This
combination of layers allows the model to learn
nuanced details essential for accurately identifying
dog breeds (Mali, Chapte, et al., 2024).
Figure 2: Model Architecture Details.
Identifying New Species of Dogs Using Machine Learning Model
197

4 RESULT AND DISCUSSION
4.1 Dataset Description
The dataset used to perform breed identification is
Stanford's Dogs Breed Identification dataset from
Kaggle (Asreddy, Shingade, et al., 2019). In this
dataset there are 120 breeds of dog with the two sets
for training and testing (Pathak, Sakore, et al., 2019).
In the training set, there are 10224 images present and
in testing also 10220 images are available (Lonari,
Jagdale, et al., 2021). We also used the label CSV file
to map with the training image dataset.
4.2 Algorithms Used
Convolutional Neural Networks (CNNs): These deep
learning models are highly effective for image
classification tasks, as they can automatically capture
and learn spatial feature hierarchies (Mali, Sawant, et
al., 2023).
Transfer Learning with Pre-trained CNNs:
Utilizing pre-trained models such as ResNet, VGG, or
Inception, which have been trained on extensive
datasets like ImageNet, can greatly enhance
performance.
Ensemble Methods: Combining predictions from
multiple CNN architectures can improve model
robustness and accuracy, providing a more reliable
identification process.
Support Vector Machines (SVM) with CNN
features: Extracting features from CNNs and
classifying those using SVMs is an alternative method
to improve accuracy, especially when computational
resources are limited.
4.3 Result of Feature Extraction
Figure 3: Model Architecture Details.
The Figure 3 presents the results of feature extraction
using different deep learning architectures on the
Stanford Dogs dataset. The dataset, along with a CSV
file containing breed labels, is utilized to evaluate each
model's capability to extract relevant features for dog
breed classification.
4.3.1 Feature Extraction Algorithms
Five prominent deep learning models—Inception,
Xception, VGG16, NasNet, and ResNet—are
employed for extracting features from the dataset.
Each model captures a unique set of image
characteristics, which are crucial for accurate
classification.
4.3.2 Feature Map Dimensions
The feature maps generated by each algorithm are
represented with dimensions (number of samples,
feature size). All models processed 9,691 samples, but
the feature size varies:
• Inception and Xception generate feature
maps of size 2,048.
• VGG16 produces a smaller feature map with a
dimension of 512.
• NasNet captures a broader range of features
with a size of 4,032.
• ResNet outputs feature maps of size 1,536.
4.3.3 Final Feature Map
The final feature map for the classification model is
shaped to (9691,9664), representing the total feature
size used in the final stage.
4.3.4 Classification Accuracy
The model utilizing these features achieves an
impressive accuracy of 98.80% post-classification,
demonstrating the effectiveness of the feature
extraction methods in distinguishing between dog
breeds.
4.3.5 Validation and Testing
The model's compilation involves selecting an
appropriate optimizer, metrics, and loss function. The
dataset is divided into training, validation, and testing
sets using the Train-Test Split method. The training set
is used to train the model, with progress closely
monitored during the process.
Model Evaluation
Once training is complete, the model is evaluated on a
separate test set, consisting of images not included in
the training process. The model's accuracy is then
INCOFT 2025 - International Conference on Futuristic Technology
198

measured based on its performance on the test set.
Predict Dog Breed
The SVM classifier that has been trained is utilized to
forecast the dog breeds in your testing dataset. The
SVM takes the CNN-extracted feature vectors as
input and provides breed predictions.
Evaluation Metrics
The performance of the hybrid CNN-SVM model is
evaluated using metrics such as accuracy and log loss.
The trained model is tested on unseen data to
evaluate its performance and make predictions. An
example image is selected for testing purposes. The
selected image is preprocessed to match up to the
model's predicted input shape. The model's
predictions are generated for the pre-processed
image. The breed label with the highest probability
from the model's output is selected as the predicted
result. Original breed label of the image is obtained
from the dataset. The original and predicted breed
labels are displayed to compare. By applying this
model to the existing image dataset from Kaggle we
can achieve an accuracy of 98.80%. Figure 4
illustrates the accuracy achieved during training and
testing validation.
Figure 4: Training and Testing Accuracy.
After training the model, predictions were made
on the testing data for breed identification. Figure 5
presents the classification report for the dog breed
identification process. The model achieved high
precision, recall, and F1-scores across most classes,
with a minor decrease in precision for class 7,
highlighting its robust performance in accurately
classifying dog breeds.
Figure 5: Project F1 Score Support.
The two confusion matrices displayed below in
Fig. 6 visualize the classification performance of a dog
breed recognition model across multiple classes.
These matrices provide insight into the model's
accuracy and areas for potential improvement:
Matrix on the Left
This matrix demonstrates a high level of classification
accuracy. The majority of predictions align perfectly
along the diagonal, indicating that the predicted breeds
closely match the actual breeds for most samples. The
diagonal's bright color represents a high count of
correct predictions, signifying that the model
consistently identifies the correct breed with minimal
errors. This suggests strong performance in
distinguishing among the various dog breeds within
the dataset.
Figure 6: Confusion Matrix.
Matrix on the Right
In contrast, the second confusion matrix highlights
areas of misclassification. While there is still a
noticeable diagonal line of correct predictions, the
presence of off-diagonal elements indicates some
misclassified cases. These misclassifications are
represented by the lighter, scattered dots outside the
diagonal. The range of colors in the scale, from dark
(low) to light (high), shows varying counts of errors in
different classes. This pattern suggests that certain
breeds might share similar features, leading to
occasional confusion by the model.
Identifying New Species of Dogs Using Machine Learning Model
199

Figure 7: Output Generation.
The matrices collectively emphasize that while
the model is effective at distinguishing most breeds,
there are specific classes where refinement is needed
to improve prediction accuracy. These visuals are
crucial for diagnosing specific instances of confusion
between similar dog breeds, guiding further
enhancements in the model's training and feature
extraction techniques.
In Figure 7, the machine learning model
accurately identifies the dog’s breed as a "Golden
Retriever." The image shows a gentle, sleepy Golden
Retriever cuddling with a pillow, illustrating the
model’s capability to recognize breeds even when the
dog isn’t in a classic, posed position. This result
highlights the model’s effectiveness in identifying
real-world images, adding to its reliability for
practical breed recognition.
Figure 8: Result.
In Fig 8, a dog, which visually looks like a dingo,
lies on sandy ground, almost merged with the earthy
tones of the background. The model is agreeing with
the ground truth and points to "dingo" as the breed.
The text overlay is showing the original label
("dingo") and the model's prediction - also "dingo".
This outcome shows successful identification by the
model, a clear result where it can differentiate the
breed in a more real-world-like setting.
5 CONCLUSIONS
This paper combines a CNN-based feature extractor
with an SVM classification algorithm to identify dog
breeds based on images. The model achieves 98.80%
accuracy in classifying a total of 120 different breeds
given the Kaggle dataset. An accuracy of this
magnitude showcases the capability of the model to
classify such different visual features of all breeds, and
it would be a reliable tool for dog breed identification.
The outcomes highlight the capability of the model to
assist areas like animal welfare, veterinary care,
behaviour training, and nutrition through the
identification of breeds accurately.
Future development of this work will extend
beyond breed identification to aim for a more
advanced application. Access to veterinary services
and breed recognition in addition to emotion detection
can be incorporated into a platform, on which pet
owners and professionals can understand the specific
needs of each breed to help guarantee better animal
welfare. Emotion detection would help interpret
behavioural signals, and a medical support feature
would provide breed-specific health information, thus
creating a versatile solution for pet care and
management.
REFERENCES
P. B. More, A. N. Jadhav, I. Khatik, S. Singh, V. K. Borate
and Y. K. Mali, “Sign Language Recognition Using
Hand Gestures”, 2024 3rd International Conference for
Advancement in Technology (ICONAT), GOA, India,
2024, pp. 1-5, doi:10.1109/ICONAT61936.20
24.10774685.
Y. Mali, M. E. Pawar, A. More, S. Shinde, V. Borate and
R. Shirbhate, “Improved Pin Entry Method to Prevent
Shoulder Surfing Attacks”, 2023 14th International
Conference on Computing Communication and
Networking Technologies (ICCCNT), Delhi, India,
2023, pp. 1-6, doi: 10.1109/ICCCNT56998.2023.
10306875.
INCOFT 2025 - International Conference on Futuristic Technology
200

Y. K. Mali and A. Mohanpurkar, “Advanced pin entry
method by resisting shoulder surfing attacks”, 2015
International Conference on Information Processing
(ICIP), Pune, India, 2015, pp. 37-42, doi:
10.1109/INFOP.2015.7489347.
S. A. Nalawade, R. Pattnaik, S. Kadam, P. P. Lodha, Y. K.
Mali and V. K. Borate, “Smart Contract System with
Block-chain Capability for Improving Supply Chain
Management”, 2024 3rd International Conference for
Advancement in Technology (ICONAT), GOA, India,
2024, pp. 1-7, doi: 10.1109/ICONAT61936.
2024.10774955.
S. P. Patil, S. Y. Zurange, A. A. Shinde, M. M. Jadhav, Y.
K. Mali and V. Borate, “Upgrading Energy Productivity
in Urban City Through Neural Support Vector Machine
Learning for Smart Grids”, 2024 15th International
Conference on Computing Communication and
Networking Technologies (ICCCNT), Kamand, India,
2024, pp. 1-5, doi: 10.1109/ICCCNT61001.2024.1
0724069.
S. Modi, M. Modi, V. Alone, A. Mohite, V. K. Borate and
Y. K. Mali, “Smart shopping trolley Using Arduino
UNO”, 2024 15th International Conference on
Computing Communication and Networking
Technologies (ICCCNT), Kamand, India, 2024, pp.1-6,
doi: 10.1109/ICCCNT61001.2024.10725524.
U. Mehta, S. Chougule, R. Mulla, V. Alone, V. K. Borate
and Y. K. Mali, “Instant Messenger Forensic System”,
2024 15th International Conference on Computing
Communication and Networking Technologies
(ICCCNT), Kamand, India, 2024, pp. 1- 6, doi:
10.1109/ICCCNT61001.2024.10724367.
P. Shimpi, B. Balinge, T. Golait, S. Parthasarathi, C. J.
Arunima and Y. Mali, “Job Crafter - The One- Stop
Placement Portal”, 2024 15th International Conference
on Computing Communication and Networking
Technologies (ICCCNT), Kamand, India, 2024, pp.1-8,
doi: 10.1109/ICCCNT61001.2024.10725010.
V. Ingale, B. Wankar, K. Jadhav, T. Adedoja, V. K. Borate
and Y. K. Mali, “Healthcare is being revolutionized by
AI-powered solutions and technological integration for
easily accessible and efficient medical care”, 2024 15th
International Conf. on Computing Communication and
Networking Technologies (ICCCNT), Kamand, India,
2024, pp. 1-6, doi: 10.1109/ICCCNT61001.202
4.10725646.
U. Mulani, V. Nandgaonkar, R. Mulla, S. Sonavane, V. K.
Borate and Y. K. Mali, “Smart Contract System with
Blockchain Capability for Improved Supply Chain
Management Traceability and Transparency”, 2024
15th International Conf. on Computing Communication
and Networking Technologies (ICCCNT), Kamand,
India, 2024, pp. 1-7, doi: 10.1109/ICCCNT610
01.2024.10723871.
S. Sonawane, U. Mulani, D. S. Gaikwad, A. Gaur, V. K.
Borate and Y. K. Mali, “Blockchain and Web3.0 based
NFT Marketplace”, 0 2024 15
th
International
Conference on Computing Communication and
Networking Technologies (ICCCNT), Kamand, India,
2024, pp. 1-6, doi: 10.1109/ICCCNT61001.2024.
10724420.
P. Mandale, S. Modi, M. M. Jadhav, S. S. Khawate, V. K.
Borate and Y. K. Mali, “Investigation of Different
Techniques on Digital Actual Frameworks Toward
Distributed Denial of Services Attack,” 2024 15th
International Conference on Computing 14
Communication and Networking Technologies
(ICCCNT), Kamand, India, 2024, pp. 1-6, doi:
10.1109/ICCCNT61001.2024.10725776.
D. Sengupta, S. A. Nalawade, L. Sharma, M. S. J. Kakade,
V. K. Borate and Y. K. Mali, “Enhancing File Security
Using Hybrid Cryptography”, 2024 15th International
Conference on Computing Communication and
Networking Technologies (ICCCNT), Kamand, India,
2024, pp. 1-8, doi: 10.1109/ICCCNT61001.2024.
10724120.
A. More, S. Khane, D. Jadhav, H. Sahoo and Y. K. Mali,
“Auto-shield: Iot based OBD Application for Car Health
Monitoring”, 2024 15th International Conf. on
Computing Communication and Networking
Technologies (ICCCNT), Kamand, India, 2024, pp.
1-10, doi: 10.1109/ICCCNT61001.2024.10726186.
U. H. Wanaskar, M. Dangore, D. Raut, R. Shirbhate, V. K.
Borate and Y. K. Mali, “A Method for Re- identifying
Subjects in Video Surveillance using Deep Neural
Network Fusion,” 2024 15th International Conference on
Computing Communication and Networking
Technologies (ICCCNT), Kamand, India, 2024, pp.1-4,
doi:10.1109/ICCCNT61001.2024.10726255.
A. More, O. L. Ramishte, S. K. Shaikh, S. Shinde and Y. K.
Mali, “Chain-Checkmate: Chess game using
blockchain,” 2024 15th International Conference on
Computing Communication and Networking
Technologies (ICCCNT), Kamand, India, 2024, pp. 1- 7,
doi: 10.1109/ICCCNT61001.2024.10725572.
J. D. Palkar, C. H. Jain, K. P. Kashinath, A. O. Vaidya, V.
K. Borate and Y. K. Mali, “Machine Learning Approach
for Human Brain Counselling”, 2024 15th International
Conference on Computing Communication and
Networking Technologies (ICCCNT), Kamand, India,
2024, pp. 1-8, doi: 10.1109/ICCCNT61
001.2024.10723852.
M. Dangore, S. Modi, S. Nalawade, U. Mehta, V. K. Borate
and Y. K. Mali, “Revolutionizing Sport Education With
AI”, 2024 15th International Conf. on Computing
Communication and Networking Technologies
(ICCCNT), Kamand, India, 2024, pp.1-8,
doi:10.1109/ICCCNT61001.2024.10724009.
M. Dangore, D. Bhaturkar, K. M. Bhale, H. M. Jadhav, V.
K. Borate and Y. K. Mali, “Applying Random Forest for
IoT Systems in Industrial Environments”, 2024 15
th
International Conference on Computing
Communication and Networking Technologies
(ICCCNT), Kamand, India, 2024, pp. 1-7, doi:
10.1109/ICCCNT61001.2024.10725751.
A. More, S. R. Shinde, P. M. Patil, D. S. Kane, Y. K. Mali
and V. K. Borate, “Advancements in Early Detection of
Lung Cancer using YOLOv7”, 2024 5th International
Conference on Smart Electronics and Communication
Identifying New Species of Dogs Using Machine Learning Model
201

(ICOSEC), Trichy, India, 2024, pp. 1739-1746,doi:
10.1109/ICOSEC61587.2024.10722534.
A. O. Vaidya, M. Dangore, V. K. Borate, N. Raut, Y. K.
Mali and A. Chaudhari, “Deep Fake Detection for
Preventing Audio and Video Frauds Using Advanced
Deep Learning Techniques,” 2024 IEEE Recent
Advances in Intelligent Computational Systems
(RAICS), Kothamangalam, Kerala, India, 2024, pp. 1-
6, doi: 10.1109/RAICS61201.2024.10689785.
Sawardekar, S., Mulla, R., Sonawane, S., Shinde, A.,
Borate, V., Mali, Y.K. (2025). Application of Modern
Tools in Web 3.0 and Blockchain to Innovate
Healthcare System. In: Rawat, S., Kumar, A., Raman,
A., Kumar, S., Pathak, P. (eds) Proceedings of Third
International Conf. on Computational Electronics for
Wireless Communications. ICCWC 2023. Lecture Notes
in Networks and Systems, vol 962. Springer, Singapore.
https://doi.org/10.1007/978-981-97-1946- 4_2
Modi, S., Mali, Y., Kotwal, R., Kisan Borate, V., Khairnar,
P., Pathan, A. (2024). Hand Gesture Recognition and
Real-Time Voice Translation for the Deaf and Dumb. In:
Jain, S., Mihindukulasooriya, N., Janev, V., Shimizu,
C.M. (eds) Semantic Intelligence. ISIC 2023. Lecture
Notes in Electrical Engineering, vol 1258. Springer,
Singapore.https://doi.org/10.1007/978-981- 97-7356-
5_35.
Bhongade, A., Dargad, S., Dixit, A., Mali, Y.K., Kumari, B.,
Shende, A. (2024). Cyber Threats in Social Metaverse
and Mitigation Techniques. In: Somani, A.K., Mundra,
A., Gupta, R.K., Bhattacharya, S., Mazumdar, A.P. (eds)
Smart Systems: Innovations in Computing. SSIC 2023.
Smart Innovation, Systems and Technologies, vol 392.
Springer, Singapore. https://doi.org/10.1007/978-981-
97-3690-4_34.
Mali, Yogesh. & TejalUpadhyay, “Fraud Detection in Online
Content Mining Relies on the Random Forest
Algorithm”, SWB 1, no. 3 (2023): 13-20.
Kale, Hrushikesh, Kartik Aswar, and Dr Yogesh Mali Kisan
Yadav, “Attendance Marking using Face Detection”,
International Journal of Advanced Research in Science,
Communication and Technology: 417-424.
Inamdar, Faizan, Dev Ojha, C. J. Ojha, and D. Y. Mali. “Job
Title Predictor System,” International Journal of
Advanced Research in Science, Communication and
Technology (2024): 457-463.
Jagdale, Sudarshan, Piyush Takale, Pranav Lonari,
Shraddha Khandre, and Yogesh Mali, “Crime
Awareness and Registration System”, International
Journal of Scientific Research in Science and
Technology 5, no. 8 (2020).
Modi, S., Mali, Y., Sharma, L., Khairnar, P., Gaikwad,
D.S., Borate, V. (2024). A Protection Approach for
Coal Miners Safety Helmet Using IoT. In: Jain, S.,
Mihindukulasooriya, N., Janev, V., Shimizu, C.M.
(eds) Semantic Intelligence. ISIC 2023. Lecture Notes
in Electrical Engineering, vol 1258. Springer,
Singapore. https://doi.org/10.1007/978-981-97-7356-
5_30.
Y. K. Mali, L. Sharma, K. Mahajan, F. Kazi, P. Kar and A.
Bhogle, “Application of CNN Algorithm on X- Ray
Images in COVID-19 Disease Prediction,” 2023 IEEE
International Carnahan Conference on Security
Technology (ICCST), Pune, India, 2023, pp. 1-6, doi:
10.1109/ICCST59048.2023.10726852.
Shabina Modi, “Automated Attendance Monitoring System
for Cattle through CCTV.”, REDVET, vol. 25, no. 1,
pp. 1025 -1034, Sep. 2024.
Y. Mali and V. Chapte, Grid based authentication system
International Journal of Advance Research in Computer
Science and Management Studies, vol. 2, no. 10, pp. 93-
99, October 2014.
Rajat Asreddy, Avinash Shingade, Niraj Vyavhare, Arjun
Rokde and Yogesh Mali, “A Survey on Secured Data
Transmission Using RSA Algorithm and
Steganography”, International Journal of Scientific
Research in Computer Science Engineering and
Information Technology (IJSRCSEIT), vol. 4, no. 8, pp.
159-162, September-October 2019, ISSN 2456–3307.
Jyoti Pathak, Neha Sakore, Rakesh Kapare, Amey Kulkarni
and Prof. Yogesh Mali, “Mobile Rescue Robot”,
International Journal of Scientific Research in Computer
Science Engineering and Information Technology
(IJSRCSEIT), vol. 4, no. 8, pp. 10-12, September-
October 2019, ISSN 2456–3307.
Pranav Lonari, Sudarshan Jagdale, Shraddha Khandre,
Piyush Takale and Prof Yogesh Mali, “Crime
Awareness and Registration System”, International
Journal of Scientific Research in Computer Science
Engineering and Information Technology (IJSRCSEIT),
vol. 8, no. 3, pp. 287-298, May-June 2021, ISSN 2456–
3307.
Yogesh Mali and Nilay Sawant, “Smart Helmet for Coal
Mining”, International Journal of Advanced Research in
Science Communication and Technology (IJARSCT),
vol. 3, no. 1, February 2023.
INCOFT 2025 - International Conference on Futuristic Technology
202