This paper presents a survey and experimental study of monocular pedestrian detection methods. The objective is to provide an overview of current state-of-the-art techniques and evaluate their performance on a large-scale benchmark dataset. The paper covers both methodological and experimental aspects of pedestrian detection, including the main components of a pedestrian detection system: hypothesis generation (ROI selection), classification (model matching), and tracking. The experimental study evaluates a diverse set of state-of-the-art systems, including wavelet-based AdaBoost cascade, HOG/linSVM, NN/LRF, and combined shape-texture detection. The dataset used for evaluation consists of 21,790 monocular images captured from a vehicle in a 27-minute urban drive, with 56,492 manual labels and 259 trajectories of fully visible pedestrians. The dataset is publicly available for benchmarking purposes. The paper discusses various approaches to pedestrian detection, including generative and discriminative models, feature extraction techniques, and classifier architectures. It also covers tracking methods, which are essential for inferring trajectory-level information. The study evaluates the performance of different detection systems under two scenarios: generic pedestrian detection and pedestrian detection from a moving vehicle. The results indicate that HOG/linSVM performs well at higher image resolutions and lower processing speeds, while the wavelet-based AdaBoost cascade is superior at lower resolutions and near real-time processing speeds. The paper also discusses the importance of using a large-scale dataset with sequential images to evaluate both hypothesis generation and tracking components of pedestrian detection systems. The study highlights the challenges of pedestrian detection, including the wide range of possible pedestrian appearances and the need for robust and efficient localization methods in both generic and application-specific settings. The paper concludes with a discussion of the results and their implications for future research in pedestrian detection.This paper presents a survey and experimental study of monocular pedestrian detection methods. The objective is to provide an overview of current state-of-the-art techniques and evaluate their performance on a large-scale benchmark dataset. The paper covers both methodological and experimental aspects of pedestrian detection, including the main components of a pedestrian detection system: hypothesis generation (ROI selection), classification (model matching), and tracking. The experimental study evaluates a diverse set of state-of-the-art systems, including wavelet-based AdaBoost cascade, HOG/linSVM, NN/LRF, and combined shape-texture detection. The dataset used for evaluation consists of 21,790 monocular images captured from a vehicle in a 27-minute urban drive, with 56,492 manual labels and 259 trajectories of fully visible pedestrians. The dataset is publicly available for benchmarking purposes. The paper discusses various approaches to pedestrian detection, including generative and discriminative models, feature extraction techniques, and classifier architectures. It also covers tracking methods, which are essential for inferring trajectory-level information. The study evaluates the performance of different detection systems under two scenarios: generic pedestrian detection and pedestrian detection from a moving vehicle. The results indicate that HOG/linSVM performs well at higher image resolutions and lower processing speeds, while the wavelet-based AdaBoost cascade is superior at lower resolutions and near real-time processing speeds. The paper also discusses the importance of using a large-scale dataset with sequential images to evaluate both hypothesis generation and tracking components of pedestrian detection systems. The study highlights the challenges of pedestrian detection, including the wide range of possible pedestrian appearances and the need for robust and efficient localization methods in both generic and application-specific settings. The paper concludes with a discussion of the results and their implications for future research in pedestrian detection.