Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Deep Learning for On-Street Parking Violation Prediction

Thien Nhan Vo

TL;DR

The paper addresses the problem of predicting illegal on-street parking rates without expensive sensor deployments by introducing a deep learning framework that uses indirect signals (weather, time, and historical data) and a novel PoI-based sector encoding to estimate hourly violation rates per sector. It combines sine-encoded temporal features, weather averages, and context indicators (holidays, pandemic) with a Gaussian data augmentation scheme to handle sparse and noisy annotations, all within a residual deep neural network trained on THESi Thessaloniki data. Key findings show the baseline MAE of 0.175 improves to 0.169 on a test set and down to 0.146 when smoothing is applied to the full dataset, demonstrating substantial predictive gains and practical viability for real-time parking guidance. The approach offers a scalable, low-cost enhancement to on-street parking systems and lays groundwork for future work in graph-based spatial modeling and richer contextual features.

Abstract

Illegal parking along with the lack of available parking spaces are among the biggest issues faced in many large cities. These issues can have a significant impact on the quality of life of citizens. On-street parking systems have been designed to this end aiming at ensuring that parking spaces will be available for the local population, while also providing easy access to parking for people visiting the city center. However, these systems are often affected by illegal parking, providing incorrect information regarding the availability of parking spaces. Even though this can be mitigated using sensors for detecting the presence of cars in various parking sectors, the cost of these implementations is usually prohibiting large. In this paper, we investigate an indirect way of predicting parking violations at a fine-grained level, equipping such parking systems with a valuable tool for providing more accurate information to citizens. To this end, we employed a Deep Learning (DL)-based model to predict fine-grained parking violation rates for on-street parking systems. Moreover, we developed a data augmentation and smoothing technique for further improving the accuracy of DL models under the presence of missing and noisy data. We demonstrate, using experiments on real data collected in Thessaloniki, Greece, that the developed system can indeed provide accurate parking violation predictions.

Deep Learning for On-Street Parking Violation Prediction

TL;DR

The paper addresses the problem of predicting illegal on-street parking rates without expensive sensor deployments by introducing a deep learning framework that uses indirect signals (weather, time, and historical data) and a novel PoI-based sector encoding to estimate hourly violation rates per sector. It combines sine-encoded temporal features, weather averages, and context indicators (holidays, pandemic) with a Gaussian data augmentation scheme to handle sparse and noisy annotations, all within a residual deep neural network trained on THESi Thessaloniki data. Key findings show the baseline MAE of 0.175 improves to 0.169 on a test set and down to 0.146 when smoothing is applied to the full dataset, demonstrating substantial predictive gains and practical viability for real-time parking guidance. The approach offers a scalable, low-cost enhancement to on-street parking systems and lays groundwork for future work in graph-based spatial modeling and richer contextual features.

Abstract

Illegal parking along with the lack of available parking spaces are among the biggest issues faced in many large cities. These issues can have a significant impact on the quality of life of citizens. On-street parking systems have been designed to this end aiming at ensuring that parking spaces will be available for the local population, while also providing easy access to parking for people visiting the city center. However, these systems are often affected by illegal parking, providing incorrect information regarding the availability of parking spaces. Even though this can be mitigated using sensors for detecting the presence of cars in various parking sectors, the cost of these implementations is usually prohibiting large. In this paper, we investigate an indirect way of predicting parking violations at a fine-grained level, equipping such parking systems with a valuable tool for providing more accurate information to citizens. To this end, we employed a Deep Learning (DL)-based model to predict fine-grained parking violation rates for on-street parking systems. Moreover, we developed a data augmentation and smoothing technique for further improving the accuracy of DL models under the presence of missing and noisy data. We demonstrate, using experiments on real data collected in Thessaloniki, Greece, that the developed system can indeed provide accurate parking violation predictions.
Paper Structure (4 sections, 6 equations, 4 figures, 1 table)

This paper contains 4 sections, 6 equations, 4 figures, 1 table.

Table of Contents

  1. Introduction
  2. Proposed Method
  3. Experiments
  4. Conclusions

Figures (4)

  • Figure 1.1: A total of 19 PoIs were defined and used for representing various sectors of the used on-street parking system
  • Figure 2.1: The proposed data augmentation and smoothing method allows for assigning observed violation rates into nearby time slots, reducing the impact of sparse annotations and missing data scans.
  • Figure 3.1: Spatial distribution of parking violations. Darker values indicate higher average parking violation rate per sector
  • Figure 3.2: Validation error during training when using two different setups and evaluation sets.