LICENSE PLATE DETECTION USING Yolov8x AND EASY
OCR
1st Pallapati Abhinairaj Please enter the name Please enter the name
Department of Computer Department of Computer Please enter the name Department of Computer
Science and Engineering Science and Engineering Department of Computer Sciences and Engineering
Amrita School of Computing Amrita School of Computing Science and Engineering College of Applied Studies
(Amrita Vishwa (Amrita Vishwa Amrita School of Computing and Community Service
Vidyapeetham) Vidyapeetham) (Amrita Vishwa (King Saud University)
Chennai, India Chennai, India Vidyapeetham) Riyadh, India
pabhinairaj@[Link]
Chennai, India
Abstract— Vehicle management and traffic surveillance applications. Overall, our research advances the practical
systems of today depend heavily on improving license plate uses of ALPR systems and contributes to the field of image
recognition. With regard to Automatic License Plate processing and deep learning.
Recognition (ALPR) systems, our research offers a novel
method utilizing EasyOCR and YOLOv8x object detection. In this study, we introduce a novel method for detecting
Effective license plate identification is made possible by and recognizing license plates, which is crucial for
YOLOv8x, which provides exceptional accuracy rates of 85 to contemporary vehicle management and traffic surveillance
90%, even in difficult low-resolution photos. With EasyOCR, systems. The YOLOv8x object detector, which is well-
we can extract text from plates and improve ALPR accuracy known for its precision and effectiveness in identifying a
by addressing the limitations of OCR. Artifacts, skewed variety of items, including license plates, is incorporated into
viewpoints, complex backdrops, and handwritten characters our system. We concentrate on real-time recognition, which
are all handled in this method. The incorporation of EasyOCR is essential for uses in law enforcement, automated toll
into ALPR elucidates the function of OCR, highlighting its collection, and vehicle tracking.
significance in enhancing the precision and ease of use of the
system. TensorFlow RT streamlines model inference for By utilizing YOLOv8x's capabilities and a large dataset,
applications requiring low latency, high throughput, and we are able to recognize license plates with an impressive
minimal memory usage, enabling seamless implementation 85–90% accuracy rate, even in difficult low-resolution
across several hardware configurations. All things considered, photos. We incorporate Optical Character identification
our approach improves ALPR's capabilities and shows how (OCR) utilizing EasyOCR to further improve identification
important OCR is to increasing the usefulness of ALPR accuracy, allowing the extraction of alphanumeric letters
systems for automated toll collection, vehicle monitoring, and from license plates. We have assembled a wide range of
law enforcement. license plate recognition photos from different sources that
have been expertly annotated. The dataset provides a
Keywords— OCR ,Yolov8x ,Deep Learning , Image balanced mix of positive and negative instances to support
Processing , ALPR system, Ultralytics , Tensor Flow Rt. effective model training. It is a useful tool for developing and
implementing algorithms since it makes evaluation more
I. INTRODUCTION efficient and avoids overfitting. A methodical methodology
A vital part of contemporary vehicle management and was utilized to gather an extensive dataset appropriate for our
traffic monitoring systems is license plate recognition. This assignment. The dataset consists of a large number of
study offers a novel method for recognizing license plates, an samples (3000 in total), each of which represents a unique
important tool for automated toll collection, vehicle license plate for a vehicle.
monitoring, and law enforcement, among other uses. Our
research attempts to offer a thorough approach for both the We can assess the algorithms' performance in real-time
identification of license plates and the extraction of instantaneous gesture recognition using a webcam or an
characters from the plates using optical character recognition giving an input video, which gives us information about their
(OCR). Our Automatic License Plate Recognition (ALPR) accuracy, efficiency, and suitability for License Plate
system, which uses the YOLOv8x object detector, achieves recognition tasks. In addition to making a significant
an excellent 85 to 90% accuracy rate in detecting license contribution to the field of plate recognition, the study's
plates, even in difficult low-resolution photos. findings have applications for maintaining and advancing
traditional License Plate detection.
We acknowledge the limitations inherent in OCR systems
and address complexities such as artifacts, distorted II. LITERATURE REVIEW
perspectives, intricate backgrounds, and handwritten A. License Plate recognition
characters. By incorporating EasyOCR into our ALPR
project, we enhance the accuracy and usefulness of ALPR The two main components of our project are EasyOCR,
systems. Additionally, TensorFlow RT facilitates the smooth which is well-known for its adaptability in optical character
deployment of our models in production contexts, ensuring recognition (OCR), and YOLOv8X, which is acclaimed for
low-latency, high-throughput, and low-memory consumption its skill in object identification. Our system uses YOLOv8X
and EasyOCR to provide reliable license plate recognition EasyOCR is a popular OCR technique that uses deep
and real-time OCR capabilities. Accurately detecting license learning algorithms to extract text from photos in real-time
plates under a variety of environmental situations is difficult. with high accuracy. Finding and deciphering alphanumeric
Our method gives individual character recognition a lower characters in the input image is the main goal of EasyOCR.
priority than strong license plate detection. The unique size A deep learning model is trained on a wide range of
of license plates allows for efficient modeling using the annotated photos with different backgrounds, typefaces, and
YOLOv7 architecture, enabling accurate detection styles as part of this procedure. Through the analysis of
unaffected by aspect ratio variations [1]. The Yolov8 is visual elements and patterns, the model gains the ability to
utilized to map key features onto identified license plates. identify and categorize individual characters.
This process results in the extraction of essential landmarks EasyOCR uses the trained model to partition the
crucial for accurate license plate recognition, including input picture into smaller areas and forecast the characters
indicators for differentiating between various types of plates. inside each region during the recognition phase. EasyOCR
delivers great accuracy and resilience in text recognition
tasks by utilizing sophisticated neural network structures and
techniques, such as recurrent neural networks (RNNs) and
convolutional neural networks (CNNs). EasyOCR's ability
to handle complicated photos with a range of lighting
conditions, angles, and distortions is one of its main
features. Comprehensive solutions for jobs like automated
license plate recognition (ALPR) are made possible by
EasyOCR's smooth interface with other system components,
such as license plate detecting algorithms.
D. Convolutional Neural Network (CNN)
Fig. 1. Identification of Number Plates using Yolov8
Convolutional Neural Networks (CNNs) have become a
mainstay in a number of computer vision applications, such
B. YOLOv8
as object identification with frameworks like YOLO, license
Modern object identification systems like You Only plate recognition, and optical character recognition (OCR).
Look Once (YOLO) are renowned for their quickness and CNNs are particularly good at extracting complex patterns
precision. It detects objects in real time by processing and characteristics from textual pictures for OCR jobs,
pictures via a convolutional neural network (CNN) in a which allows for precise character identification [3]. Large
single pass. The input image is divided into a grid by datasets of labeled characters may be used to train them to
YOLO, which then forecasts bounding boxes and class acquire comprehensive representations, enabling them to
probabilities for every grid cell. recognize characters with accuracy across a variety of fonts,
YOLO is available in several versions, each with sizes, and styles.
unique enhancements and adjustments. The idea of creating Furthermore, in license plate identification systems,
a grid out of the image and forecasting bounding boxes and CNNs are essential for character recognition. CNNs are
class probabilities was first presented by YOLOv1. By capable of accurately classifying and interpreting characters
adding features like batch normalization and anchor boxes to found on license plates by utilizing deep learning
increase accuracy and speed, YOLOv2 improves upon this. architectures. This makes it possible for cars to be accurately
YOLOv3 included a feature pyramid network and multiple identified and tracked in a variety of traffic surveillance and
scale bounding box prediction, greatly optimizing the management applications [2].
architecture. A number of enhancements were brought about All things considered, CNNs are extremely powerful at
by YOLOv4, including the utilization of the CSPDarknet53 processing complicated visual input, which is why they are
backbone and sophisticated data augmentation methods to so important for tasks like OCR, license plate identification,
increase speed. Ultralytics' YOLOv5 featured a streamlined and character recognition. This allows for the development
design and used fewer parameters to get excellent results. of automated systems that are useful for a wide range of
Yolov7 algorithms can be used to recognize and track real-world applications.
objects as they move through a production line, allowing for
more efficient and accurate manufacturing. The next E. Scaling
latest one is YOLOv8, which is also originates with To enhance the effectiveness of machine learning models
Ultralytics. Like YOLOv7, it hasn't been officially released such as YOLOv8 and EasyOCR, data scaling in license plate
yet, but it's anticipated to improve upon YOLOv5's identification entails preprocessing procedures to normalize
advantages by perhaps providing even greater effectiveness the input data, guaranteeing homogeneity.
and performance.
The latest one is YOLOv9 presented a number of
enhancements, such as the utilization of the CSPDarknet53
backbone and sophisticated data augmentation methods to
increase performance.
C. Object Character Recognition(OCR)
One common method for obtaining text from pictures or
documents is optical character recognition, or OCR.
III. RELATED WORK uses the AdaBoost cascade classifier for character
Recognition of vehicle number Plate, with each paper segmentation and a pixel-level two-class classification
contributing to this domain in its unique way as shown some approach for random-position object recognition.
of them in “Table I”. "Automatic License Plate Detection Comparing experimental results on Macau license plates to
and Recognition System for Security Purposes" by Debajit state-of-the-art methods, performance is better, indicating
Sarma, Anindita Bora and Atul Bhagat published in June possible regional or national application. It achieves
2023, this paper presents an ALPR system including accuracy up to 98 percent.[5]. M.A. Jawale a, P. William a,
detection, segmentation, and character recognition A.B. Pawar b and Nikhil Marriwala c developed a Plate
components for secure areas. It correctly recognizes recognition system , In addition to highlighting deep
segmented digits using a CNN classifier, which is essential learning innovations and IoT sensor integration, this
for access control and law enforcement [1]. The second provides an overview of ALPR methodology and suggests a
paper, "AI-based license plate detection of on-road vehicle," full system with distinct approaches for every job, such as
authored by Henakshi Das, Prasant Kumar Sahu and Kukil character recognition utilizing CNN, MobileNet, Inception
Khanikar and published in July 2023, this technique uses V3, and ResNet 50[6].
mobile cameras and sophisticated recognition technologies IV. METHODOLOGY AND IMPLEMENTATION
to overcome problems with vehicle license plate detection.
The system highlights the usage of SSD MobileNet and A. Design phase
EasyOCR for effective recognition, achieving accuracy rates The research uses deep learning models and computer
of 86% to 95% and demonstrating the potential of mobile vision techniques to identify license plates in photos of
devices in ANPR applications [2]. "License Plate Detection vehicles. In order to construct deep learning methods,
and Recognition: A Study of Review" by Naaman PyTorch and torch vision are used in conjunction with
Omar,Subhi R. [Link] , Mohammed [Link] and OpenCV for image processing. Furthermore, important
Rizgar R. Zebari, This work investigates ALP detection and libraries like NumPy and Matplotlib are used for data
recognition using machine learning and image processing visualization and manipulation, respectively.
methods. It goes over pre-processing techniques like as The architectural diagram for License Plate Detection is
contrast augmentation and Gaussian filtering and shown in "Fig. 2", which provides an overview of the
emphasizes how ALP technology has advanced with CNN methodical structure used in this procedure.
models and deep semantic segmentation networks. The
studies reviewed here contribute to the development of B. Implementation Phase
techniques and methodologies for automating license plate The programming language utilized for this project is
detection and recognition, crucial for traffic surveillance and Python. To facilitate the design process, the Visual Studio
civic management systems. In 2022, Tian Xia, Ruizhuo source code editor was employed. To develop a model for
Zhang, Yongjun Zhang and Baisong Chen proposed a recognition of detecting Plates of vehicles, multiple phases
detection model for automatic recognition of License Plate are necessary, namely:
using Yolov7 [4]. Later, A Review on License Plate
Recognition using Hybrid Broad Learning System with a
fully convolutional network by C. L. Philip Chen and
Bingshu Wang turned out to a new fusion which combines a
wide learning system with a fully convolutional network. It
TABLE I. PREVIOUS WORK
Year Title Author(s) Methodology
2023 Automatic License Plate Detection and Debajit Sarma, Anindita Bora With CNN-based character identification,
Recognition System for Security Purposes and Atul Bhagat [1]. segmentation, and detection capabilities, this
study suggests an effective automated
license plate recognition system.
2022 AI-based license plate detection of on-road Henakshi Das, Prasant Kumar This approach is done by taking pictures
vehicle Sahu, Kukil Khanikar [2] with mobile cameras, then utilizing
EasyOCR and SSD MobileNet to recognize
license plates.
2017 License Plate Detection and Recognition: A Naaman Omar,Subhi R. M. Dataset created from internet images for
Study of Review Zeebaree , Mohammed A. training and testing, recognition using HOG
[Link] and Rizgar R. features and SVM classifier
Zebari [3]
2013 Application of YOLOv7 and Transformer Tian Xia, Ruizhuo Zhang, YOLOv7 is trained and tested using a
structures to small object (license plate) Yongjun Zhang and Baisong particular dataset as part of the technique for
detection in complex traffic scenes Chen [4] complicated traffic situations involving
license plate identification.
Fig. 2. Proposed Architecture
criteria, including accuracy, speed, and suitability for
1. Data Collection and Preparation the project specifications, were used to evaluate each
design. Following a thorough analysis, YOLOv8 was
Our project's license plate detection and identification
dataset was selected from a variety of sources, such as
publicly accessible databases, internet archives, and
specialized data gathering initiatives. To be more precise,
our dataset, which we acquired from the Roboflow platform
via the "car_license_plates" project inside the "moin"
workspace, includes a variety of vehicle makes and model
names. This method made sure that our dataset included all
of the details of license plates in a variety of real-world
situations, giving us a solid base on which to train and assess determined to be the best option because of its capacity
our model. to strike a balance between accuracy and real-time
Every image in the dataset was painstakingly annotated performance.
before training, with bounding boxes used to indicate the
locations of license plates. Throughout the collection,
precise and consistent annotations were made possible using
annotation tools such as LabelImg and VGG Image
Annotator. Through the use of these annotations as ground
truth labels during model training, the model was able to
efficiently pick up on the spatial properties of license plates.
We used a number of preprocessing methods to improve the
variety and quality of the dataset.
. To achieve robust performance, our license plate
detection and identification model makes use of a wide
range of datasets, careful annotation, and thorough Fig. 3. Dataset Classification
preprocessing. Below is the Fig 3 showing the division of
Dataset in our
The training setup was created using YOLOv8x as the
2. Model Selection and Training: preferred architecture as shown in “Fig.3”. This included
setting up the model configuration parameters, software
To find the best model for the job of detecting and dependencies, and hardware resources such GPUs for
recognizing license plates, a thorough assessment of acceleration in the training environment. In order to
maximize model convergence and speed up the training
many object detection architectures was the first step in process, GPU acceleration was essential. The training setup
the model selection and training process. A number of was created using YOLOv8x as the preferred architecture as
shown in “Fig.3”.
Fig. 3. License Plate training samples
3. Evaluation Metrics and Validation
The training setup was created using YOLOv8x as the
preferred architecture as shown in “Fig.3”. This included Following training, the model's performance was
setting up the model configuration parameters, software assessed using a variety of measures to see how well it could
dependencies, and hardware resources such GPUs for reliably identify and recognize license plates. Important
acceleration in the training environment. In order to information regarding the trained model, such as its
maximize model convergence and speed up the training architecture, parameters, and computational complexity,
process, GPU acceleration was essential. The model's were disclosed in the model summary. With more than 25
performance was adjusted by extensive hyperparameter million parameters and 218 layers, the model showed a
tweaking. Through testing and validation, parameters like significant ability to identify complex patterns in the data.
learning rate, batch size, and number of training epochs were With an inference speed of 10.7 milliseconds per picture, the
repeatedly changed. The goal of this iterative procedure was model's computational efficiency was also emphasized,
to maximize the model's capacity to reliably identify and guaranteeing real-time performance in deployment
recognize license plates in a variety of environmental scenarios.
settings.
A crucial part of the procedure was integrating the
chosen dataset into the training pipeline. The model was
trained using a dataset that was carefully annotated with
bounding boxes around license plates and collected from a
variety of sources. During training, data loaders were put up
to quickly process and feed large batches of photos into the
model as shown in the “Fig.2”, making efficient use of the
dataset. Moreover, optimizing the loss function was essential
to improving the performance of the model. In order to
minimize false positives and negatives and promote accurate
license plate localization and classification, the loss function
was modified. The goal of this optimization procedure was
to raise the model's precision in identifying and detecting
license plates in practical situations. The main procedures
and factors involved in training a license plate detection and
identification model utilizing YOLOv8 architecture are
highlighted in this methodical approach, which offers a clear
picture of the model selection and training phase.
Fig. 4. Training model Script
In order to objectively assess the trained model's
performance, it underwent rigorous validation against an
independent dataset throughout the assessment phase.
Evaluation measures were calculated to measure the model's
performance in properly detecting license plates in a variety
of environmental situations. These metrics included
precision (P), recall (R), and mean average precision (mAP).
With a mAP50 of 0.887, a recall of 0.765, and an accuracy
of 0.997, the model demonstrated encouraging performance
as shown in the confusion matrix in the “Fig.5”. These
numbers show how well the model locates and categorizes
license plates inside photos.
Fig. 5. Confusion Matrix
4. Inference and Real-World Deployment
The trained model must be integrated with systems or apps
that can interpret live or recorded video streams involving
automobiles in order to be used for inference in real-world
settings. The approach is usually implemented on edge
devices, including dedicated hardware or embedded
systems, so that real-time performance may be guaranteed
without depending on distant servers. The model in this
deployment arrangement examines every frame of the video
stream in order to precisely identify and detect license
plates. Various environmental elements, including as
illumination, weather, and vehicle orientations, affect the
model's performance in real-world circumstances.
In conclusion, the trained model must be integrated
into edge devices for real-time video stream processing in
order to be used for inference in real-world situations. Its
resilience and performance are revealed by extensive testing
in a variety of environmental situations, which informs
continuous developments to guarantee accurate license plate
detection and identification in a range of real-world
scenarios.
5. Performance Analysis and Comparison
Following a discussion of the model's robustness, accuracy,
and speed in relation to current techniques or benchmarks,
you can add the confusion matrix as an extra performance
parameter. A more thorough examination of the model's
advantages and disadvantages is made possible by the
confusion matrix, which provides a thorough breakdown of
the classification outcomes. Readers may better grasp the
model's classification performance by seeing the confusion
matrix shown with other performance measures like
accuracy, recall, and F1 score is shown in the “Fig.6”. They During the last step, the existing model was tested by
can see how successfully the algorithm differentiates detecting License Plate and displaying the Number Plate text
between various classes and spot any trends or patterns in on the screen. This approach involves using the OCR library
the incorrect classifications. Furthermore, you may gain to capture License Plate with the camera and the YOLO
important insights into your model's relative performance by framework to recognize License Plates. The framework helps
contrasting its confusion matrix with those of other aggregate coordinates of the License Plate using Object
approaches or benchmarks. You may draw attention to Detection.
specific instances in which your model performs particularly The current model is then used to preview the precise
well or poorly, allowing for a more thorough comparison Text
and identifying possible areas for development. being
You may think about adding the confusion matrix given
as an appendix or supplemental material for readers who based
would like to look at it in further depth in addition to on these
presenting it in the main text. By doing this, the confusion
matrix is kept easily available without interfering with the
primary story as shown below in “Fig.6”.
coordinates, resulting in real-time detection of License Plate.
The images below namely “Fig. 7”, “Fig. 8”, “Fig. 9”,
“Fig.
10”,
and
“Fig.
11”
Fig. 6. Performance Analysis
V. RESULT depicts some of the recognized License plate. Almost all
License Plates are easily recognized, except for some that
A. Testing Scenario
may be more difficult to notice, such as Low Lighting,
Extreme angles and Partial occlusions etc. like.
Additionally, even if they are almost identical, some License
Plate with similar coordinate points.
Our automated number plate Detection and recognition
model was assessed both directly in real-world settings and
through the use of a specific library called Easy-OCR. While
accuracy was measured using library testing, real-time
performance was evaluated by direct testing. With the use of
this dual strategy, a thorough review was guaranteed,
Fig. 7. Vehicle showing text “L656-XH”
offering quantitative performance measurements and
insights into real-world application.
B. Result while using Library
Our evaluation's findings show how effective our license
plate detection and identification model is in practical
situations. Our model can reliably recognize license plates
with high accuracy, as seen by its precision, recall, and mean
average precision (mAP) values of 0.997, 0.765, and 0.887
respectively. By demonstrating the model's capacity to
discriminate between true positives, false positives, true
negatives, and false negatives, the confusion matrix analysis
provides further insight into the model's classification Fig. 8. Vehicle showing text “H-644-LX”
performance.
Result of Direct Testing using Demo Video
Fig. 9. Vehicle showing text “K-884-RS” [2] Henakshi Das, Prasant Kumar Sahu, Kukil Khanikar, "AI-based
License Plate Detection of On-road Vehicles," presented at the 2023
International Conference on Microwave, Optical, and Communication
Engineering (ICMOCE), Bhubaneswar, India, May 26-28, 2023. DOI:
10.1109/ICMOCE57812.2023.10167115
[3] Naaman Omar, Subhi R. M. Zeebaree, Mohammed A. M. Sadeeq,
Rizgar R. Zebari, "License Plate Detection and Recognition: A Study
of Review," presented at the 2nd International Conference of
Mathematics, Applied Sciences, Information and Communication
Technology, January 2023. AIP Conference Proceedings. DOI:
10.1063/5.
[4] Tian Xia, Ruizhuo Zhang, Yongjun Zhang, Baisong Chen,
"Application of YOLOv7 and Transformer Structures to Small Object
(License Plate) Detection in Complex Traffic Scenes," presented at
the 2022 4th International Conference on Machine Learning, Big Data
Fig. 10. Vehicle showing text “66-HH-07” and Business Intelligence (MLBDBI), Shanghai, China, October 28-
30, 2022. DOI: 10.1109/MLBDBI58171.2022.00031
[5] C. L. Philip Chen, Bingshu Wang, "Random-Positioned License Plate
Recognition Using Hybrid Broad Learning System and Convolutional
Networks," published in 2020, presents a novel method combining
wide learning systems and convolutional networks for license plate
recognition.
[6] M.A. Jawale, P. William, A.B. Pawar, Nikhil Marriwala,
"Implementation of Number Plate Detection System for Vehicle
Registration Using IoT and Recognition Using CNN," published in
Volume 27, June 2023, 100761, focuses on automatic license plate
recognition (ALPR) crucial for intelligent transportation systems. The
study emphasizes deep learning techniques and IoT sensors,
comprising character recognition, image pre-processing, plate
extraction, and segmentation, employing CNN, MobileNet, Inception
V3, and ResNet 50.
Fig. 11. Vehicle showing text “L6o5-HZ”
[7] Muhammad Ali, Peshawa & Faraj, Rezhna. (2014). Data
Normalization and Standardization: A Technical Report.
VI. CONCLUSION 10.13140/RG.2.2.28948.04489.
[8] Vasudeva, Ruchita., H.R., Mamatha. (2023). Recognition of Indian
With the help of YOLOv8x and Easy-OCR, our project Classical Dance Hand Gestures. 814-820. doi:
has effectively created a reliable system for automatic 10.1109/ICICT57646.2023.10134484.
license plate recognition and detection. We trained and [9] D., S., Arpitha., M., Balasubrahmanyam., D., Anil, Kumar. (2022).
assessed our model using an extensive dataset that included Depth based Indian Classical Dance Mudra's Recognition using
Support Vector Machine. 885-888. doi:
a range of license plate kinds and ambient circumstances. 10.1109/ICSSIT53264.2022.9716486.
We achieved this by utilizing YOLOv8x's sophisticated [10] K., V., V., Kumar., P.V.V., Kishore. (2017). Indian Classical Dance
object identification skills and Easy-OCR's optical character Mudra Classification Using HOG Features and SVM Classifier.
recognition technologies. International Journal of Electrical and Computer Engineering,
7(5):2537-2546. doi: 10.1007/978-981-10-5544-7_65.
Our model showed excellent accuracy in identifying and
[11] V. Adithya, P. R. Vinod and U. Gopalakrishnan, "Artificial neural
detecting license plates in various lighting situations, angles, network based method for Indian sign language recognition," 2013
and occlusions through direct testing in real-world scenarios. IEEE Conference on Information & Communication Technologies,
Even with certain difficult cases—like dim illumination or Thuckalay, India, 2013, pp. 1080-1085, doi:
partial occlusions—the overall performance held up well, 10.1109/CICT.2013.6558259.
demonstrating the efficacy of our strategy. Accurate [12] [Link]
recognition
extraction of alphanumeric characters from detected license
[13] Devi, Mampi & Saharia, Sarat & Bhattacharyya, Dhruba K. (2015).
plates was made possible by integration with Easy-OCR, Dance Gesture Recognition: A Survey. International Journal of
which improved recognition accuracy even more. Our Computer Applications. 122. 19-26. 10.5120/21696-4803.
system's overall usefulness and dependability were greatly [14] Ankita, Bisht., Riya, Bora., Goutam, Saini., Pushkar, Shukla. (2017).
enhanced by this connection, making it appropriate for real- Indian Dance Form Recognition from Videos. 123-128. doi:
10.1109/SITIS.2017.
world uses in a variety of fields like traffic monitoring,
[15] MediaPipe. (n.d.). MediaPipe Hands. mediapipe.
parking management, and law enforcement. [Link] python-
solution-api.
Our study demonstrates how cutting-edge object detection [16] S K, Sriram & Sinha, Nishant. (2021). Gestop: Customizable Gesture
methods and cutting-edge OCR technology may be used to Control of Computer Systems. [Link]
create effective and dependable automatic license plate [17] [Link]
identification systems. In order to guarantee the system's [18] Divya, Hariharan., Tinku, Acharya., Sushmita, Mitra. (2011).
Recognizing hand gestures of a dancer. 186-192. doi: 10.1007/978-3-
efficacy and scalability in practical deployments, further 642-21786-9_32
efforts will concentrate on improving performance, fine- [19] M., Reshma., Balakrishnan, Kannan., V, P, Jagathy, Raj., S., Shailesh.
tuning the model, and resolving any outstanding issues. (2023). Cultural heritage preservation through dance digitization: A
review. Digital Applications in Archaeology and Cultural Heritage,
28:00257-00257. doi: 10.1016/[Link].2023.e00257
REFERENCE [20] Soumitra, Samanta., Pulak, Purkait., Bhabatosh, Chanda. (2012).
[1] Debajit Sarma, Anindita Bora, Atul Bhagat, "Automatic License Plate Indian Classical Dance classification by learning dance pose bases.
Detection and Recognition System for Security Purposes," presented at 265-270. doi: 10.1109/WACV.2012.6163050|
2023 IEEE Guwahati Subsection Conference (GCON), Guwahati, [21] Mampi, Devi., Sarat, Saharia. (2016). A two-level classification
India, June 23-25, 2023. DOI: 10.1109/GCON58516.2023.10183638 scheme for single-hand gestures of Sattriya dance. doi:
10.1109/ICADW.2016.7942540
[22] Nurfitri, Anbarsanti., Ary, Setijadi, Prihatmanto. (2014). Dance
modelling, learning and recognition system of aceh traditional dance
based on hidden Markov model. 78:86-92. doi: 10.11113/JT.V78.693