Implementasi Arsitektur InceptionResNet-v2 dan Squared Hinge Loss  (Studi Kasus Klasifikasi Pose Yoga)

MUHAMMAD ICHWAN; ANNISA OLGA ZERLINDA

doi:10.26760/mindjournal.v7i2.124-138

Implementasi Arsitektur InceptionResNet-v2 dan Squared Hinge Loss (Studi Kasus Klasifikasi Pose Yoga)

MUHAMMAD ICHWAN, ANNISA OLGA ZERLINDA

Sari

ABSTRAK

Aplikasi computer vision dapat digunakan untuk mengurangi kecelakaan yang terjadi ketika melakukan yoga akibat postur tubuh saat melakukan yoga yang tidak tepat. Pada penelitian, dilakukan sebagai langkah awal dalam penentuan model klasifikasi yang digunakan pada aplikasi yang dapat melakukan koreksi postur tubuh saat yoga. Penelitian dilakukan dengan mengklasifikasikan 11 studi kasus pose yoga dengan mengimplementasikan arsitektur InceptionResnet-v2 dan Squared Hinge Loss. Pada hasil yang didapatkan, model dengan kinerja terbaik diperoleh pada learning rate 0.0001, epochÂ 200, scaling residual InceptionResnet blok A 0.15, blok B 0.1, dan blok C 0.2, serta blok InceptionResnet 5 iterasi, B 10 iterasi, dan C 5 iterasi dengan kinerja arsitektur berdasarkan hasil evaluasi performa model didapatkan 89.98% accuracy, 90.38% precision, 89.79% recall, 89.83% F1 score, and 99% specificity pada pengujian 888 data uji dengan 11 pose yoga berbeda. Rata-rata pengujian waktu klasifikasi 1.301s dan loss 0.9494.

Kata kunci: CNN, InceptionResnet-v2, Klasifikasi Citra, Squared Hinge Loss, Yoga

ABSTRACT

Computer vision applications can be used to reduce accidents caused by improper posture while doing yoga. In this study, it was carried out as a first step in determining the classification model used in applications that can make corrections to a body posture while doing yoga. The research was conducted to classify 11 case studies of yoga poses by implementing the InceptionResnet-v2 architecture and Squared Hinge Loss. In the results, the model with the best performance was obtained at a learning rate of 0.0001, epoch 200, scaling residual InceptionResnet block A 0.15, block B 0.1, and block C 0.2, and 5 iteration of InceptionResnet block A, 10 iterations of block B, and 5 iterations of block C. The results of the model performance evaluation obtained 89.98% accuracy, 90.38% precision, 89.79% recall, 89.83% F1 score, and 99% specificity in the test of 888 test data with 11 different yoga poses and 1.301s average testing time of the classification model and loss 0.9494.

Keywords: CNN, Image Classification, InceptionResnet-v2, Squared Hinge Loss, Yoga

Teks Lengkap:

PDF

Referensi

Agarap, A. F. (2017). An Architecture Combining Convolutional Neural Network (CNN) and Support Vector Machine (SVM) for Image Classification. 5â€“8. http://arxiv.org/abs/1712.03541

Agarwal, M., Gupta, S. K., & Biswas, K. K. (2020). Development of Efficient CNN model for Tomato crop disease identification. Sustainable Computing: Informatics and Systems, 28, 100407. https://doi.org/10.1016/j.suscom.2020.100407

Al-masni, M. A., Kim, D. H., & Kim, T. S. (2020). Multiple skin lesions diagnostics via integrated deep convolutional networks for segmentation and classification. Computer Methods and Programs in Biomedicine, 190, 105351. https://doi.org/10.1016/j.cmpb.2020.105351

Dong, N., Zhao, L., Wu, C. H., & Chang, J. F. (2020). Inception v3 based cervical cell classification combined with artificially extracted features. Applied Soft Computing, 93, 106311. https://doi.org/10.1016/j.asoc.2020.106311

He, K., Zhang, X., Ren, S., & Sun, J. (2016). Deep residual learning for image recognition. Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition, 2016-Decem, 770â€“778. https://doi.org/10.1109/CVPR.2016.90

Jose, J., & Shailesh, S. (2021). Yoga Asana Identification: A Deep Learning Approach. IOP Conference Series: Materials Science and Engineering, 1110(1), 012002. https://doi.org/10.1088/1757-899x/1110/1/012002

Kumar, D., & Sinha, A. (2020). Yoga Pose Detection and Classification Using Deep Learning. International Journal of Scientific Research in Computer Science, Engineering and Information Technology, 160â€“184. https://doi.org/10.32628/cseit206623

Nguyen Quoc, T., & Truong Hoang, V. (2020). Medicinal Plant identification in the wild by using CNN. International Conference on ICT Convergence, 2020-Octob, 25â€“29. https://doi.org/10.1109/ICTC49870.2020.9289480

Peng, S., Huang, H., Chen, W., Zhang, L., & Fang, W. (2020). More trainable inception-ResNet for face recognition. Neurocomputing, 411, 9â€“19. https://doi.org/10.1016/j.neucom.2020.05.022

Rishan, F., Lanka, S., Nijabdeen, S., Lanka, S., Silva, B. De, Lanka, S., Rupasinghe, L., Lanka, S., Alawathugoda, S., Lanka, S., Liyanapathirana, C., & Lanka, S. (2020). Infinity Yoga Tutor : Yoga Posture.

Szegedy, C., Ioffe, S., Vanhoucke, V., & Alemi, A. A. (2017). Inception-v4, inception-ResNet and the impact of residual connections on learning. 31st AAAI Conference on Artificial Intelligence, AAAI 2017, 4278â€“4284.

Tang, Y. (2013). Deep Learning using Linear Support Vector Machines. http://arxiv.org/abs/1306.0239

Verma, M., Kumawat, S., Nakashima, Y., & Raman, S. (2020). Yoga-82: A new dataset for fine-grained classification of human poses. IEEE Computer Society Conference on Computer Vision and Pattern Recognition Workshops, 2020-June, 4472â€“4479. https://doi.org/10.1109/CVPRW50498.2020.00527

DOI: https://doi.org/10.26760/mindjournal.v7i2.124-138