The Smart in Smart Cities: A Framework for Image Classification Using Deep Learning

The need for a smart city is more pressing today due to the recent pandemic, lockouts, climate changes, population growth, and limitations on availability/access to natural resources. However, these challenges can be better faced with the utilization of new technologies. The zoning design of smart cities can mitigate these challenges. It identifies the main components of a new smart city and then proposes a general framework for designing a smart city that tackles these elements. Then, we propose a technology-driven model to support this framework. A mapping between the proposed general framework and the proposed technology model is then introduced. To highlight the importance and usefulness of the proposed framework, we designed and implemented a smart image handling system targeted at non-technical personnel. The high cost, security, and inconvenience issues may limit the cities' abilities to adopt such solutions. Therefore, this work also proposes to design and implement a generalized image processing model using deep learning. The proposed model accepts images from users, then performs self-tuning operations to select the best deep network, and finally produces the required insights without any human intervention. This helps in automating the decision-making process without the need for a specialized data scientist.

Automated summary

The need for a smart city has become more urgent in recent years due to various factors such as pandemics, lockdowns, climate change, population growth, and limitations on natural resources. This article proposes a general framework for designing a smart city and introduces a technology-driven model to support it. By designing and implementing a smart image handling system and a generalized image processing model using deep learning, the importance and practicality of this framework are highlighted.