Estimating Passenger Density in Trains through Crowd Counting Modeling
Dublin Core
Title
Estimating Passenger Density in Trains through Crowd Counting Modeling
Subject
Commuter Line, Density Estimation, Crowd Counting, P2PNet, Individual Localization
Description
The Greater Jakarta Commuter Rail, also known as the KRL Commuter Line, is one of the primary
transportation choices for many people due to its comfort and efficiency. However, the level of user
dissatisfaction is still relatively high, particularly regarding the frequent and unpredictable overcrowding of
trains. To address this issue, our research develops an Artificial Intelligence-based model to predict train
passenger density through crowd counting. By utilizing the proposed k-F1 metric By utilizing the proposed
k-F1 metric, which balances the impact of False Positives and False Negatives in crowd density predictions by measuring the proximity of predicted points to the nearest ground truth within a scaled threshold and a constructed dataset of train density, we compare three object detection approaches: bounding box prediction (YOLOv5), density map (CSRNet), and proposal point (P2PNet). In our experiments, YOLOv5 surpassed
other models in performance, achieving a Mean Absolute Error (MAE) of 1.41 and a k-F1 score of 0.91, while maintaining a fast inference speed of 300 milliseconds per frame. This model’s strength lies in scenarios with fewer people and larger objects, such as passengers, within the frame. Conversely, P2PNet and CSRNet were less successful under these conditions, achieving MAEs of 3.49 and 4.98, and k-F1 scores of 0.77 and 0.35 respectively. However, it is important to note that P2PNet and CSRNet are better suited for denser and more congested environments, such as peak hours or at major transit hubs, where trains typically experience high crowd densities. The proposed density estimation method can be applied to real-time image-based CCTV systems to predict train congestion and facilitate transportation management decisions.
transportation choices for many people due to its comfort and efficiency. However, the level of user
dissatisfaction is still relatively high, particularly regarding the frequent and unpredictable overcrowding of
trains. To address this issue, our research develops an Artificial Intelligence-based model to predict train
passenger density through crowd counting. By utilizing the proposed k-F1 metric By utilizing the proposed
k-F1 metric, which balances the impact of False Positives and False Negatives in crowd density predictions by measuring the proximity of predicted points to the nearest ground truth within a scaled threshold and a constructed dataset of train density, we compare three object detection approaches: bounding box prediction (YOLOv5), density map (CSRNet), and proposal point (P2PNet). In our experiments, YOLOv5 surpassed
other models in performance, achieving a Mean Absolute Error (MAE) of 1.41 and a k-F1 score of 0.91, while maintaining a fast inference speed of 300 milliseconds per frame. This model’s strength lies in scenarios with fewer people and larger objects, such as passengers, within the frame. Conversely, P2PNet and CSRNet were less successful under these conditions, achieving MAEs of 3.49 and 4.98, and k-F1 scores of 0.77 and 0.35 respectively. However, it is important to note that P2PNet and CSRNet are better suited for denser and more congested environments, such as peak hours or at major transit hubs, where trains typically experience high crowd densities. The proposed density estimation method can be applied to real-time image-based CCTV systems to predict train congestion and facilitate transportation management decisions.
Creator
Bryan Tjandra, Oey Joshua Jodrian, Nyoo Steven Christopher Handoko, Alfan Farizki Wicaksono
Source
http://dx.doi.org/10.21609/jiki.v18i1.1314
Publisher
Faculty of Computer Science Universitas Indonesia
Date
2025-02-08
Contributor
Sri Wahyuni
Rights
e-ISSN : 2502-9274 printed ISSN : 2088-7051
Format
PDF
Language
English
Type
Text
Coverage
Jurnal Ilmu Komputer dan Informasi (Journal of Computer Science and Information)
Files
Collection
Citation
Bryan Tjandra, Oey Joshua Jodrian, Nyoo Steven Christopher Handoko, Alfan Farizki Wicaksono, “Estimating Passenger Density in Trains through Crowd Counting Modeling,” Repository Horizon University Indonesia, accessed May 23, 2025, https://repository.horizon.ac.id/items/show/8940.