 The YOLO-V5MS model is a smart city solution that enhances pedestrian detection and security measures by optimizing video stream acquisition, replacing the original backbone convolution with RepBlock, incorporating a bio-inspired squeeze NIC Citation module, using K-Means algorithm and Rednecks image augmentation during training, and adopting the focal AU approach for loss computation. The model achieves a 96.5% MAV value and a 21.3% increase in average inference speed on an internally developed smart city dataset, making it suitable for performing pedestrian detection within an Internet of over 50 video surveillance cameras. This article was authored by Fong Jing Song and Pung Lee.