YOLOv5-Based Cigarette Detection and Audible Warning System for Public Health and Environmental Sustainability

Nur Kuban Torun; Oğuzhan Tatli; Merve Önalan

Makale detayları

YOLOv5-Based Cigarette Detection and Audible Warning System for Public Health and Environmental Sustainability

Yazarlar:

Nur Kuban Torun | Oğuzhan Tatli | Merve Önalan

Yayıncı:

Gazi University

Ana Başlıklar:

Sürdürülebilir Kalkınma ve Kamu Yararına Bilgi Sistemleri

Dergi:

Gazi Üniversitesi Fen Bilimleri Dergisi Part C: Tasarım ve Teknoloji Vol.13 Issue 2, pp.746-762

Belge Tipi:

Makale

Yayın Tarihi:

6/2025

Belge Dili:

Türkçe

Document ID (Org No.):

2001289897

pISSN:

2147-9526

Anahtar kelimeler:

Özet:

 The use of cigarettes causes numerous health issues worldwide, leading to millions of deaths annually. To mitigate the individual and societal impacts of smoking, including passive smoking, smoke-free zone initiatives play a crucial role in protecting public health. In this study, the YOLOv5 model was utilized for real-time detection of smoking behavior. A dataset comprising 3,912 images captured under varying lighting, angle, and background conditions was created and augmented to 10,176 images using various data augmentation techniques. The data were labeled on the Roboflow platform and trained using the Ultralytics HUB integrated with the Google Colab Pro environment. During training, the YOLOv5s model achieved a 90.6 precision rate, 73.4 recall, and a 0.825 mAP0.5. Detected smoking behavior was communicated to users via an auditory warning system. This system played a pre-recorded sound file using the Pygame library whenever a predefined confidence threshold was exceeded. Thus, compliance with smoke-free zone regulations was supported, and the risk of passive smoking was minimized. This study provides valuable insights into protecting human health and improving environmental air quality, thereby making a meaningful contribution to the literature. This study contributes to the Sustainable Development Goal of Good Health and Well-Being by offering an innovative technology that serves the public interest. It provides valuable outcomes for human health and environmental air quality, making a meaningful contribution to the literature.

 The use of cigarettes causes numerous health issues worldwide, leading to millions of deaths annually. To mitigate the individual and societal impacts of smoking, including passive smoking, smoke-free zone initiatives play a crucial role in protecting public health. In this study, the YOLOv5 model was utilized for real-time detection of smoking behavior. A dataset comprising 3,912 images captured under varying lighting, angle, and background conditions was created and augmented to 10,176 images using various data augmentation techniques. The data were labeled on the Roboflow platform and trained using the Ultralytics HUB integrated with the Google Colab Pro environment. During training, the YOLOv5s model achieved a 90.6 precision rate, 73.4 recall, and a 0.825 mAP0.5. Detected smoking behavior was communicated to users via an auditory warning system. This system played a pre-recorded sound file using the Pygame library whenever a predefined confidence threshold was exceeded. Thus, compliance with smoke-free zone regulations was supported, and the risk of passive smoking was minimized. This study provides valuable insights into protecting human health and improving environmental air quality, thereby making a meaningful contribution to the literature. This study contributes to the Sustainable Development Goal of Good Health and Well-Being by offering an innovative technology that serves the public interest. It provides valuable outcomes for human health and environmental air quality, making a meaningful contribution to the literature.

 The use of cigarettes causes numerous health issues worldwide, leading to millions of deaths annually. To mitigate the individual and societal impacts of smoking, including passive smoking, smoke-free zone initiatives play a crucial role in protecting public health. In this study, the YOLOv5 model was utilized for real-time detection of smoking behavior. A dataset comprising 3,912 images captured under varying lighting, angle, and background conditions was created and augmented to 10,176 images using various data augmentation techniques. The data were labeled on the Roboflow platform and trained using the Ultralytics HUB integrated with the Google Colab Pro environment. During training, the YOLOv5s model achieved a 90.6 precision rate, 73.4 recall, and a 0.825 mAP0.5. Detected smoking behavior was communicated to users via an auditory warning system. This system played a pre-recorded sound file using the Pygame library whenever a predefined confidence threshold was exceeded. Thus, compliance with smoke-free zone regulations was supported, and the risk of passive smoking was minimized. This study provides valuable insights into protecting human health and improving environmental air quality, thereby making a meaningful contribution to the literature. This study contributes to the Sustainable Development Goal of Good Health and Well-Being by offering an innovative technology that serves the public interest. It provides valuable outcomes for human health and environmental air quality, making a meaningful contribution to the literature.

 Sigara kullanımı, dünya genelinde çok sayıda sağlık sorununa yol açarak her yıl milyonlarca insanın ölümüne neden olmaktadır. Pasif içicilik de dahil olmak üzere, sigaranın bireysel ve toplumsal etkilerini azaltmak amacıyla geliştirilen dumansız hava sahası uygulamalarının halk sağlığını korumada önemli bir rolü vardır. Bu çalışmada, YOLOv5 modeli kullanılarak sigara içme davranışının gerçek zamanlı tespiti hedeflenmiştir. Farklı ışık, açı ve arka plan koşullarında çekilen 3.912 görüntüden oluşan bir veri seti oluşturulmuş ve çeşitli veri artırma teknikleriyle bu sayı 10.176ya çıkarılmıştır. Veriler Roboflow platformunda etiketlenmiş ve Ultralytics HUB kullanılarak Google Colab Pro ortamında model eğitimi gerçekleşleştirilmiştir. YOLOv5s modeli, eğitim sürecinde 90.6 doğruluk oranı Precision, 73.4 duyarlılık Recall ve 0.825 mAP0.5 değerlerine ulaşmıştır. Tespit edilen sigara içme davranışı sesli bir uyarı sistemiyle kullanıcıya iletilmiştir. Bu sistem, güven eşiği aşıldığında önceden hazırlanmış bir ses dosyasını Pygame kütüphanesi aracılığı ile çalmaktadır. Böylece bireylerin yasaklı alanlarda yasağa uyum sağlamaları desteklenmiş ve pasif içicilik riski en aza indirilmeye çalışılmıştır. Bu çalışma, sürdürülebilir kalkınma hedeflerinden Sağlıklı ve Kaliteli Yaşam hedefine katkıda bulunarak, kamu yararına hizmet eden yenilikçi bir teknoloji sunmaktadır. İnsan sağlığı ve çevresel hava kalitesine yönelik değerli sonuçlar ortaya koyarak literatüre anlamlı bir katkı sağlamaktadır.

Makale detayları

YOLOv5-Based Cigarette Detection and Audible Warning System for Public Health and Environmental Sustainability

Yazarlar:

Nur Kuban Torun | Oğuzhan Tatli | Merve Önalan

Yayıncı:

Gazi University

Ana Başlıklar:

Sürdürülebilir Kalkınma ve Kamu Yararına Bilgi Sistemleri

Dergi:

Gazi Üniversitesi Fen Bilimleri Dergisi Part C: Tasarım ve Teknoloji Vol.13 Issue 2, pp.746-762

Belge Tipi:

Makale

Yayın Tarihi:

6/2025

Belge Dili:

Türkçe

Document ID (Org No.):

2001289897

pISSN:

2147-9526

Good Health and Well-Being

Deep Learning

Image Processing

Cigarette Detection

Sound AlertSağlık ve Kaliteli Yaşam

Derin Öğrenme

Görüntü İşleme

Sigara Tespiti

Sesli Uyarı

Sayı

Makale detayları

Yazarlar: Nur Kuban Torun | Oğuzhan Tatli | Merve Önalan

Yayıncı: Gazi University

Ana Başlıklar: Sürdürülebilir Kalkınma ve Kamu Yararına Bilgi Sistemleri

Dergi: Gazi Üniversitesi Fen Bilimleri Dergisi Part C: Tasarım ve Teknoloji Vol.13 Issue 2, pp.746-762

Belge Tipi: Makale

Yayın Tarihi: 6/2025

Belge Dili: Türkçe

Document ID (Org No.): 2001289897

pISSN: 2147-9526

Sayı

Yazarlar:

Nur Kuban Torun | Oğuzhan Tatli | Merve Önalan

Yayıncı:

Gazi University

Ana Başlıklar:

Sürdürülebilir Kalkınma ve Kamu Yararına Bilgi Sistemleri

Dergi:

Gazi Üniversitesi Fen Bilimleri Dergisi Part C: Tasarım ve Teknoloji Vol.13 Issue 2, pp.746-762

Belge Tipi:

Makale

Yayın Tarihi:

6/2025

Belge Dili:

Türkçe

Document ID (Org No.):

2001289897

pISSN:

2147-9526