You are on page 1of 18

SARANATHAN COLLEGE OF ENGINEERING – Trichy 12.

Department of Computer Science and Engineering

VISION BASED YOGA POSTURE


DETECTION AND CLASSIFICATION USING
DEEP LEARNING

TEAM MEMBERS:
GUIDE
DEEPIKA K 813818104018
GRITYKA S R 813818104024 Dr.R.SENTHAMIL SELVI M.E.,Phd.,
HARSHINI S 813818104028
KAVISWETHA L 813818104039
SARANATHAN COLLEGE OF ENGINEERING

PROBLEM STATEMENT
• The project aims to create a classifier that will have the
capability to identify yoga poses .
• The final result will be the name of the yoga and an audio
message playing the same.
• Solves the need for having a trainer to monitor the pose of a
self-learner.

CSA005- VISION BASED YOGA POSTURE DETECTION AND CLASSIFICATION USING DEEP LEARNING
2
SARANATHAN COLLEGE OF ENGINEERING

ABSTRACT

• The system takes an input from a web camera process the


frames .
• Movenet is used for detection .
• Multilayer perceptron is used for classification .
• Helps self-learners to prevent improper stances and its
consequences.

CSA005-VISION BASED YOGA POSTURE DETECTION AND CLASSIFICATION USING DEEP LEARNING 3
SARANATHAN COLLEGE OF ENGINEERING

INTRODUCTION

• Yoga is recently gaining popularity dur modern world stress


• All individuals who are willing to learn on their own could not
afford a personal trainer
• So we proposed a system that could solve this issue by
enabling a computerised version of a trainer
• Improves their stance and prevent consequences like muscle
pains and cramps.

CSA005-VISION BASED YOGA POSTURE DETECTION AND CLASSIFICATION USING DEEP LEARNING 4
SARANATHAN COLLEGE OF ENGINEERING

LITERATURE REVIEW
S.NO AUTHOR/YEAR TITLE TECHNIQUES USED
1. Vivek Anand Thoutam Yoga pose estimation and Focus on preprocessing rather
et. al /2022 feedback generation using than detection,but has good
deep learning acuracy score.

2. Josvin Jose et. al /2021 Yoga Asana Transfer learning withVGG16


identification: A deep architecture and pretrained
Learning approach ImageNet weights along with a
DNN classifier is used

3. Yuvin Wu et. al A computer vision based Grades the yoga pose image
/2021 yoga pose grading viathe lear ned contrastive
approach skeleton feature representations
but not a realtime system

4. Yash Agarwal et. al Implementation of TF pose estimation algorithm


/2020 machine learning is proposed to detect the joints
technique for and random forest classifier
identification of yoga (machine learning).
5
posesDETECTION AND CLASSIFICATION USING DEEP LEARNING
CSA005-VISION BASED YOGA POSTURE
SARANATHAN COLLEGE OF ENGINEERING

EXISTING SYSTEM

• The existing system uses machine learning packages ML5


and P5.

• It classifies only three yoga postures or classes.

• It does not mention the accuracy of the output received.

• The other existing system expects the user to upload a


picture of him/her taking the stands and then compares it
with the trainer’s picture to point-out the corrections. Hence
it is not in real-time.
CSA005-VISION BASED YOGA POSTURE DETECTION AND CLASSIFICATION USING DEEP LEARNING
6
SARANATHAN COLLEGE OF ENGINEERING

PROPOSED SYSTEM
• In our system, we used a recently established model called
Movenet. It was developed by TensorFlow.
• So, this Movenet model is used to detect the keypoints
(joints) in the human body which are then fed into our
classifier to classify as one of the seven poses mentioned as
below
1.Adhomukha svanasana (dog pose) 6.Virabhadrasana (warrior pose)
2.Vriksasanal (tree pose) 7. Balasana (child's pose)
3. Phalakasana (plank pose)
4. Bhujangasana (cobra pose)
5. Utkata Konasana (goddess pose)

CSA005-VISION BASED YOGA POSTURE DETECTION AND CLASSIFICATION USING DEEP LEARNING
7
SARANATHAN COLLEGE OF ENGINEERING

SYSTEM ARCHITECTURE

CSA005-VISION BASED YOGA POSTURE DETECTION AND CLASSIFICATION USING DEEP LEARNING
8
SARANATHAN COLLEGE OF ENGINEERING

LIST OF MODULES
1. Capturing and preprocessing
2. Detection
3. Classification

1. CAPTURING AND PREPROCESSING:


– The input for this system will be taken from the web camera.
– That input will be converted into frames and then preprocessed.

CSA005-VISION BASED YOGA POSTURE DETECTION AND CLASSIFICATION USING DEEP LEARNING 9
SARANATHAN COLLEGE OF ENGINEERING

2.DETECTION

– In this module, The preprocessed image will be given as an input


to a pre-trained model named “Movenet”.
– It gives the x and y co-ordinates of the key points along with the
confidence score .
– It has two variants:
*Lightning
*Thunder
– We decided to use the lightning model for the speed as it a real
time detection

CSA005-VISION BASED YOGA POSTURE DETECTION AND CLASSIFICATION USING DEEP LEARNING
10
SARANATHAN COLLEGE OF ENGINEERING

MOVENET MODEL
• A convolutional neural network model that runs on RGB
images and predicts human joint locations of a single person.
• The model is designed to be run in the browser using
Tensorflow.js or on devices using TF Lite in real-time,
targeting movement/fitness activities.

CSA005-VISION BASED YOGA POSTURE DETECTION AND CLASSIFICATION USING DEEP LEARNING 11
SARANATHAN COLLEGE OF ENGINEERING

3.CLASSIFICATION
– The detected co-ordinates along with the confidence score
will be given as an input to the classifier.

– Our classification model is a multi-layered perceptron


consisting of 7 layers .

– This model gives how accurate a human is towards a pose


after classification.

CSA005-VISION BASED YOGA POSTURE DETECTION AND CLASSIFICATION USING DEEP LEARNING 12
SARANATHAN COLLEGE OF ENGINEERING

PERFORMANCE OUTPUT SCREENSHOTS


TRAINING ACCURACY MODEL LOSS CURVE

TESTING ACCURACY MODEL ACCURACY

CSA005-VISION BASED YOGA POSTURE DETECTION AND CLASSIFICATION USING DEEP LEARNING 13
SARANATHAN COLLEGE OF ENGINEERING

CONCLUSIONS

• A web app with detection and classification model residing in


the backend that can identify 7 yoga poses.
• Tell how accurate a stance is .
• Audio message (name of the pose) is played for user in
difficult stance to know the result

CSA005-VISION BASED YOGA POSTURE DETECTION AND CLASSIFICATION USING DEEP LEARNING 14
SARANATHAN COLLEGE OF ENGINEERING

FUTURE ENHANCEMENT

• More number of yoga poses detection with updated model

• Multiperson detection for family and group of self

• Profiles for users to keep track of their sessions

CSA005-VISION BASED YOGA POSTURE DETECTION AND CLASSIFICATION USING DEEP LEARNING
15
SARANATHAN COLLEGE OF ENGINEERING

REFERENCES

• Deep Markup, Uranus Sin ha , "Yoga Pose Detection and Classification


Using Deep Learning", International Journal of Scientific Research in
Computer Science, Engineering and Information Technology
(IJSRCSEIT), ISSN : 2456-3307, Volume 6 Issue 6, pp. 160-184,
November-December 2020.

• Ian Gregory, Samuel Brahmaputra Olajuwon, "Implementation of


Computer Vision in Detecting Human Poses", Information
Management and Technology (Technician) 2020 International
Conference on, pp. 271-276, 2020.

CSA005-VISION BASED YOGA POSTURE DETECTION AND CLASSIFICATION USING DEEP LEARNING
16
SARANATHAN COLLEGE OF ENGINEERING

• A. Richard, O. Dali, O. Ramada and D. Ambuscade, "Yong-Guru: Real-


Time Yoga Pose Correction System Using Deep Learning Methods,"
2021 International Conference on Communication information and
Computing Technology
(ICCICT),2021,pp.16,dpi:10.1109/ICCICT50803.2021.950997

• X. Huang ET AL., "Intelligent Yoga Coaching System Based on


Posture Recognition," 2021 International Conference on Culture-
oriented Science & Technology (ICCST), 2021, pp. 290-293, dpi:
10.1109/ICCST53801.2021.00068.

• Y. Allegra, Y. Shah and A. Dharma, "Implementation of Machine


Learning Technique for Identification of Yoga Poses," 2020 IEEE 9th
International Conference on Communication Systems and Network
Technologies (CSNT), 2020, pp. 40-43, dpi:
10.1109/CSNT48778.2020.9115758.
CSA005-VISION BASED YOGA POSTURE DETECTION AND CLASSIFICATION USING DEEP LEARNING
17
SARANATHAN COLLEGE OF ENGINEERING

• F. Irishman, B. De Silva, S. Alawathugoda, S. Nijabdeen, L.


Rupasinghe and C. Liyanapathirana, "Infinity Yoga Tutor: Yoga
Posture Detection and Correction System," 2020 5th International
Conference on Information Technology Research (ICITR), 2020, pp.
1-6, doi: 10.1109/ICITR51448.2020.9310832.

• Y. -H. Lo, C. -C. Yang, H. Ho and S. -W. Sun, "richYoga: An


Interactive Yoga Recognition System Based on Rich Skeletal Joints,"
2021 IEEE International Conference on Artificial Intelligence and
Virtual Reality (AIVR), 2021,pp. 256-257, doi:
10.1109/AIVR52153.2021.00058.

• M. Gochoo et al., "Novel IoT-Based Privacy-Preserving Yoga Posture


Recognition System Using Low-Resolution Infrared Sensors and
Deep Learning," in IEEE Internet of Things Journal, vol. 6, no. 4, pp.
7192-7200, Aug. 2019, doi: 10.1109/JIOT.2019.2915095.
CSA005-VISION BASED YOGA POSTURE DETECTION AND CLASSIFICATION USING DEEP LEARNING
18

You might also like