Pedestrian Detection Using Convolutional Neural Networks: A Performance Evaluation with Metrics and Loss Analysis

Especially in autonomous driving and surveillance systems, pedestrian detection is a fundamental feature of computer vision applications. This work investigates the use of a convolutional neural network (CNN) for pedestrian identification by means of performance analysis based on important criteria including accuracy, precision, recall, F1 score, loss. With an accuracy of 99.8% and a precision of 100%, the model proves dependability in spotting areas not accessible by foot. The study emphasizes the model’s shortcomings in generalizing to unseen data, nevertheless, given a recall of 99.7% and clear over fitting shown from the loss and accuracy curves. Using data augmentation, modern designs, and regularizing methods is advised in order to meet these obstacles. The results lay a basis for more investigation meant to improve the strength and efficiency of pedestrian detecting systems.