Deep learning-based road damage detection and classification for multiple countries

Arya, Deeksha; Maeda, Hiroya; Ghosh, S. K.; Toshniwal, Durga; Mráz, Alexander; Kashiyama, Takehiro; Sekimoto, Yoshihide

doi:10.1016/j.autcon.2021.103935

Cited by 167 publications

(81 citation statements)

References 28 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Furthermore, they also use Poisson blending [26] to place the generated damages to the existing images to make the artificial patches look more natural to its containers. In another work, Arya et al, [3] suggest that the current techniques are not transferable from one country to another. Therefore, such a model is needed to save data collection, data labeling, and training time.…”

Section: B Deep Learning Based Road Damage Detectionmentioning

confidence: 99%

“…While collecting data can be done efficiently using mobile devices with GPS and camera [2], the labeling process takes time, and the detection/classification results are still limited. Also, the models learned from the data coming from one country often are not generalized enough to be transferable to different countries [3]. Additionally, providing bounding boxes and labels for road damages is error-prone and demands a massive amount of human labor to have accurate results.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Road Damage Detection and Classification with Detectron2 and Faster R-CNN

Pham

Dang

2020

Preprint

View full text Add to dashboard Cite

The road is vital for many aspects of life, and road maintenance is crucial for human safety. One of the critical tasks to allow timely repair of road damages is to quickly and efficiently detect and classify them. This work details the strategies and experiments evaluated for these tasks. Specifically, we evaluate Detectron2's implementation of Faster R-CNN using different base models and configurations. We also experiment with these approaches using the Global Road Damage Detection Challenge 2020, A Track in the IEEE Big Data 2020 Big Data Cup Challenge dataset. The results show that the X101-FPN base model for Faster R-CNN with Detectron2's default configurations are efficient and general enough to be transferable to different countries in this challenge. This approach results in F1 scores of 51.0% and 51.4% for the test1 and test2 sets of the challenge, respectively. Though the visualizations show good prediction results, the F1 scores are low. Therefore, we also evaluate the prediction results against the existing annotations and discover some discrepancies. Thus, we also suggest strategies to improve the labeling process for this dataset.

show abstract

Section: B Deep Learning Based Road Damage Detectionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Road Damage Detection and Classification with Detectron2 and Faster R-CNN

Pham

Dang

2020

Preprint

View full text Add to dashboard Cite

show abstract

“…Despite road networks' critical functions as catalysts for economic development, many governments still rely on either relatively inefficient and inaccurate human visual inspections or relatively expensive and difficult to scale laser-and high definition camera-based systems to carry out surface-level quality evaluations of asphalt roadways in order to identify potentially hazardous pavement distresses such as potholes and cracks that may cause road accidents and endanger motorists [4]. For instance as the majority of U.S. Department of Transportation (DOT) state agencies employ government workers or third party contractors to provide annual or biennial estimates of state highway pavement deterioration measuring the prevalence and severity of cracking, patching, faulting and joint deterioration per roadmile for different sections of state highways in order to meet federal reporting requirements mandated under the 2012 Moving Ahead for Progress in the 21st Century Act (MAP-21), inspectors will typically complete these estimates through on-the-ground or windshield visual surveys of road distresses in a way that therefore subjects these calculations to a large element of human error [4]- [6]. Other methods, such as driving specialized vehicles equipped with various sensors such as laser scanners [7], ground penetrating radar ("GPR") antennas [8] and high definition cameras [6] [9] such as shown in Figure 1 along sections of roadway in order to collect high-definition images and 3D reconstructions of the road pavement, are similarly impractical given the significant capital and labor costs required in operating these technologies [4].…”

Section: Introductionmentioning

confidence: 99%

A Computer Vision-assisted Approach to Automated Real-Time Road Infrastructure Management

Heitzmann¹

2022

Preprint

View full text Add to dashboard Cite

Accurate automated detection of road pavement distresses is critical for the timely identification and repair of potentially accident-inducing road hazards such as potholes and other surface-level asphalt cracks. Deployment of such a system would be further advantageous in low-resource environments where lack of government funding for infrastructure maintenance typically entails heightened risks of potentially fatal vehicular road accidents as a result of inadequate and infrequent manual inspection of road systems for road hazards.To remedy this, a recent research initiative organized by the Institute of Electrical and Electronics Engineers ("IEEE") as part of their 2020 Global Road Damage Detection ("GRDC") Challenge published in May 2020 a novel 21,041 annotated image dataset of various road distresses calling upon academic and other researchers to submit innovative deep learning-based solutions to these road hazard detection problems. Making use of this dataset, we propose a supervised object detection approach leveraging You Only Look Once ("YOLO") and the Faster R-CNN frameworks to detect and classify road distresses in real-time via a vehicle dashboard-mounted smartphone camera, producing 0.68 F1-score experimental results ranking in the top 5 of 121 teams that entered this challenge as of December 2021.

show abstract

“…Recently, various sensing methods based on image processing (e.g., images, videos, and lidar) have been developed to identify potential environmental barriers [ 33 , 34 , 35 ]. Image-based approaches collect data related to the interaction of pedestrians with the built environment.…”

Section: Introductionmentioning

confidence: 99%

The Feasibility of Information-Entropy-Based Behavioral Analysis for Detecting Environmental Barriers

Lee

Hwang

Kim

2021

IJERPH

View full text Add to dashboard Cite

The enhancement of physical activity is highly correlated with the conditions of the built environment. Walking is considered to be a fundamental daily physical activity, which requires an appropriate environment. Therefore, the barriers of the built environment should be identified and addressed. Barriers can act as external stimuli for pedestrians, so pedestrians may diversely respond to them. Based on this consideration, this study examines the feasibility of information-entropy-based behavioral analysis for the detection of environmental barriers. The physical responses of pedestrians were collected using an inertial measurement unit (IMU) sensor in a smartphone. After the acquired data were converted to behavioral probability distributions, the information entropy of each grid cell was calculated. The grid cells whereby the participants indicated that environmental barriers were present yielded relatively high information entropy values. The findings of this study will facilitate the design of more pedestrian-friendly environments and the development of diverse approaches that utilize citizens for monitoring the built environment.

show abstract

Deep learning-based road damage detection and classification for multiple countries

Cited by 167 publications

References 28 publications

Road Damage Detection and Classification with Detectron2 and Faster R-CNN

Road Damage Detection and Classification with Detectron2 and Faster R-CNN

A Computer Vision-assisted Approach to Automated Real-Time Road Infrastructure Management

The Feasibility of Information-Entropy-Based Behavioral Analysis for Detecting Environmental Barriers

Contact Info

Product

Resources

About