Crested Porcupine Optimizer for Pedestrian Detection using Deep Learning

Abstract

Pedestrian detection is a promising field of surveillance, intelligent transport systems, and smart city applications. Accurate detection of pedestrians ensures the safety of the individuals. This paper proposes an improved object detection model termed CPO-FRCNN using the Faster Region-Based Convolutional Neural Network (Faster RCNN) model with Resnet50 as the backbone architecture which is used for feature extraction. This novelty approach lies in the integration of the Crested Porcupine Optimizer (CPO), a metaheuristic optimization technique used to fine-tune the key hyperparameters such as learning rate and batch size. The proposed model reaches the convergence fast and produces improved detection performance through this optimizer-driven tuning process. An experiment result shows that notable improvement in the mean Average Precision (mAP) of the proposed model. Particularly, the CPO- FRCNN model exhibits strong detection for large-sized pedestrians and achieves higher precision and recall metrics. This study shows that the incorporation of CPO enhances the robustness of the pedestrian detection framework, which is suitable for complex and dense urban environments.