Research Paper: SmartTrash AI-Enabled Waste Detection
For code access please get in touch with me at mjzeolla4@gmail.com.
Waste management has been a growing issue for many years, with each country employing different individualistic solutions. However, many of the solutions applied are not enough to handle the large demand, and as the human population grows, the amount of trash generated will only increase. Therefore, it is imperative that practical and effective solutions be developed to challenge and address waste management. One domain space that has major implications in waste management is deep learning, specifically focusing on detection and classification. Using deep learning the team aimed to produce effective and accurate models by implementing state-of-the-art artificial intelligence architectures, ResNet, VGG, and YOLO. This paper focuses on training, testing, and optimizing each of these architectures and their respective models, in hopes of producing one capable of accurately classifying trash. The team focused on classifying trash accurately, optimizing each model with common best practices. Ultimately, the results from the experiments proved that deep learning can play a crucial role in waste management, with models being capable of classifying trash with over 85% real-world accuracy. Relying on these ResNet, VGG, and YOLO models will allow for other consumer-facing applications, such as cleaning robots, to integrate with the real world, and take action.