Privacy-aware crowd counting methods for real-world environment

Crowd counting with CCTV images is a challenging task that has gained attention due to its potential applications in safety management, urban planning, traffic congestion management, and advertising. However, privacy concerns and novelty problems make developing accurate and effective models for open-world scenarios difficult. This paper reviews and evaluates the five best-performing models for indirect and direct methods in privacy-aware crowd counting. They categorize CNN-based crowd counting methods and describe their advantages and disadvantages. The results show that Restormer, followed by CANet, produces the best results in the indirect method, while IIM performs best in the direct method. The paper concludes by discussing the key findings and their implications and highlights the need for further research to develop more effective and efficient models that can handle the unique and challenging aspects of the data and scenarios encountered in crowd counting with CCTV images.