Analisa Pengaruh Ukuran Testing Data dan Data Augmentation pada Tingkat Akurasi Deteksi Pemakaian Masker oleh Pengemudi Kendaraan menggunakan Deep Learning

Authors

  • Alief Wikarta Departemen Teknik Mesin, Institut Teknologi Sepuluh Nopember, Surabaya
  • Is Bunyamin Suryo Departemen Teknik Mesin, Institut Teknologi Sepuluh Nopember, Surabaya
  • M Khoirul Effendi Departemen Teknik Mesin, Institut Teknologi Sepuluh Nopember, Surabaya

Keywords:

deteksi masker, deep learning, pengemudi kendaraan, testing data, data augmentation

Abstract

Saat vaksin belum ditemukan dan efektifitas obat-obatan belum teruji, maka masker berperan penting untuk mengurangi transmisi virus. Untuk mendeteksi pemakaian masker dapat dilakukan dengan metode computer vision menggunakan deep learning. Pada penelitian ini dilakukan analisa pengaruh ukuran testing data pada tingkat akurasi deteksi pemakaian masker. Ada 5 macam ukuran test data yang digunakan, yaitu: 0.1, 0.15, 0.2, 0.25, 0.3. Untuk mengatur ukuran testing data menggunakan library sklearn, sementara untuk training deep learning memanfaatkan library Tensor Flow dan Keras pada bahasa pemrograman Python. Sementara itu, arsitektur deep learning yang digunakan adalah MobileNet dan MobileNetV2 dengan tambahan data augmentation. Beberapa teknik data augmentation yang digunakan adalah rotation, zoom, shear, shifting, dan horizontal flip. Hasil eksperimen menunjukkan, untuk ukuran testing data dari 0.1, 0.15, 0.2, 0.25, dan 0.3 secara berurutan didapatkan tingkat akurasi sebagai berikut: 0.9833, 0.97, 0.9923, 0.9853, dan 0.985. Hasil ini menunjukkan bahwa ukuran testing data 0.2 memiliki tingkat akurasi yang lebih baik dibandingkan dengan ukuran lainnya. Sementara itu, untuk tingkat akurasi dengan penambahan data augmentation didapatkan sebesar 0.9923 dibandingkan dengan 0.9654 yang tanpa data augmentation. Hasil eksperimen ini menunjukkan bahwa dengan data augmentation dapat meningkatkan nilai akurasi dari model deep learning.

 

When a vaccine has not been found, and the effectiveness of drugs has not been tested, face-masks play an important role in reducing virus transmission. To detect the use of face-masks can be done with the computer vision method using deep learning. In this study analyse the testing data size effect on the detection accuracy level of mask-wearing. There are 5 types of testing data sizes used, namely: 0.1, 0.15, 0.2, 0.25, 0.3. To adjust the testing data size using the Sklearn library, while for deep learning training, it uses the Tensor Flow and Keras library in the Python programming language. Meanwhile, the deep learning architectures used is MobileNet and MobileNetV2, with additional data augmentation. Some of the data augmentation techniques used are rotation, zoom, shear, shifting, and horizontal flip. The experimental results show, for the testing data sizes of 0.1, 0.15, 0.2, 0.25, and 0.3, the following levels of accuracy are obtained: 0.9833, 0.97, 0.9923, 0.9853, and 0.985 respectively. These results indicate that the 0.2 testing data size has a better accuracy level than the other sizes. Meanwhile, for the level of accuracy with the addition of data augmentation was obtained at 0.9923 compared to 0.9654 without data augmentation. This experiment shows that data augmentation can increase the accuracy value of the deep learning model.

Downloads

Download data is not yet available.

References

T. Greenhalgh, M. B. Schmid, T. Czypionka, D. Bassler, and L. Gruer, “Face masks for the public during the covid-19 crisis,” BMJ, vol. 369, Apr. 2020, doi: 10.1136/bmj.m1435.

World Health Organization, “Advice on the use of masks in the context of COVID-19. Interim guidance,” Geneva, Switzerland: WHO, 2020.

V. C.-C. Cheng et al., “The role of community-wide wearing of face mask for control of coronavirus disease 2019 (COVID-19) epidemic due to SARS-CoV-2,” Journal of Infection, Apr. 2020, doi: 10.1016/j.jinf.2020.04.024.

“Uber Introduces AI to Ensure Drivers Wear Face Masks.” https://www.securitymagazine.com/articles/92397-uber-introduces-ai-to-ensure-drivers-wear-face-masks (accessed May 17, 2020).

“Uber tells all drivers to wear face masks,” BBC News, May 14, 2020.

A. G. Howard et al., “MobileNets: Efficient Convolutional Neural Networks for Mobile Vision Applications,” arXiv:1704.04861 [cs], Apr. 2017, Accessed: Oct. 20, 2020. [Online]. Available: http://arxiv.org/abs/1704.04861.

M. Sandler, A. Howard, M. Zhu, A. Zhmoginov, and L.-C. Chen, “MobileNetV2: Inverted Residuals and Linear Bottlenecks,” arXiv:1801.04381 [cs], Mar. 2019, Accessed: Oct. 20, 2020. [Online]. Available: http://arxiv.org/abs/1801.04381.

F. Chollet, Deep Learning with Python, 1st edition. Shelter Island, New York: Manning Publications, 2017.

“prajnasb/observations,” GitHub. https://github.com/prajnasb/observations (accessed Nov. 12, 2020).

Published

08-01-2021
Counter
Abstract viewed 83 times
pdf (Bahasa Indonesia) viewed 112 times

How to Cite

[1]
A. . Wikarta, I. B. . Suryo, and M. K. . Effendi, “Analisa Pengaruh Ukuran Testing Data dan Data Augmentation pada Tingkat Akurasi Deteksi Pemakaian Masker oleh Pengemudi Kendaraan menggunakan Deep Learning”, SENTER, pp. 20–24, Jan. 2021.