Smoker Detection [Image] classification Dataset

This dataset contains 1,120 high-quality images, equally divided into two categories: Smoking (Smokers) and NotSmoking (Non-Smokers). Each class consists of 560 images curated through extensive web searches using diverse keywords such as cigarette smoking, smoker, coughing, taking inhaler, person on the phone, drinking water, and more. These keywords helped ensure a wide variety of images, making the dataset versatile and suitable for robust machine learning training.

Key Features

1. Diverse Class Representation

   • The Smoking class includes images of smokers captured from various angles and performing different gestures, offering realistic variability.
   • The NotSmoking class consists of images of non-smokers with gestures that mimic smoking actions, such as drinking water, holding a phone, using an inhaler, or coughing. This introduces intentional inter-class confusion to improve model robustness.

2. Preprocessing and Resolution

   • All images have been preprocessed and resized to a consistent resolution of 250×250 pixels, ensuring compatibility with various deep learning frameworks.

3. Training and Testing Split

   • The dataset is divided into 80% for training and validation and 20% for testing, allowing researchers to develop and evaluate deep learning models effectively.

Applications

• Deep Learning for Smoking Detection: Develop AI models for identifying smokers in images.
• Smart City Surveillance Systems: Monitor and enforce no-smoking zones in public areas.
• Environmental and Health Monitoring: Aid in campaigns for green environments and health-focused initiatives.
• Behavior Analysis Research: Study human gestures and behaviors in real-world scenarios.