Road Damage Detection Using RetinaNet

Ale, Laha; Zhang, Ning; Li, Longzhuang

doi:10.1109/bigdata.2018.8622025

Cited by 75 publications

(44 citation statements)

References 7 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…On the other hand, recent advances in generic object detection and classification algorithms [18] have now made possible to implement very sophisticated and resource efficient DL algorithms in constrained devices, such as mobile phones and smart cameras, and some of the recent implementations have started to emerge in various areas and in particular in the field of road inspection and assessment. In this sense, it is also possible now to carry out the deployment phase (acquisition and detection of structural damages) in real time using inexpensive mobile devices as well [20][21], either for individual inspections in situ or mounted on a car, making it an attractive alternative to the expensive MMS platforms.…”

Section: Related Workmentioning

confidence: 99%

“…Road damage types in the dataset proposed in [21] and their definitions 1 The codes in Table 1 are used by the Japanese Government to classify road damages and it is used by the authors to categorize structural damages in their work. This dataset is the largest currently available and has been extensively used to implement road inspection systems using deep learning architectures, such as SSD like RetinaNet [20] and MobileNet [21]. Although this dataset is relatively large, some classes or damage instances are poorly represented, such as the potholes class, which stems from the fact that these damages are quickly repaired in Japan.…”

Section: Utilized Datasetmentioning

confidence: 99%

“…In the case of RetinaNet, and as discussed in [20], this model can be trained with different backbone neural networks for the feature extraction phase (RestNet or VGG in its various configurations), trading accuracy versus inference time, obtaining in general better results than other popular Single Shot Detectors (SSD) such as YOLO, whilst yielding models amenable for embedded or mobile implementations. For instance, RetinaNet with VGG19 as backbone requires 115.7MB and can achieve inferences of 0.5s, fast enough for most applications in this domain.…”

Section: Trainingmentioning

confidence: 99%

See 2 more Smart Citations

Road Damage Detection Acquisition System Based on Deep Neural Networks for Physical Asset Management

Angulo¹,

Vega-Fernández²,

Aguilar-Lobo³

et al. 2019

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Research on damage detection of road surfaces using image processing techniques has been actively conducted, achieving considerably high detection accuracies. So far, most studies focused only on the detection of the presence or absence of damages; however, in real-world scenarios, road managers from need to clearly understand the type of damage and its extent in order to take effective action in advance or to allocate the necessary resources. Moreover, currently there are few uniform and openly available road damage datasets available, leading to a lack of a common benchmark for road damage detection. Such dataset could be used for a great variety of applications; herein, it is intended to serve as the acquisition component of a physical asset management tool which can aid governments agencies for planning purposes, or by infrastructure maintenance companies, so they can implement predictive maintenance procedures.In this paper, we make two contributions to address these issues. First, we present a large-scale road damage dataset, which includes a more balanced and representative set of damages not present in previous studies. This dataset is composed of 18,0345 road damage images captured with a smartphone installed on a car, with 45,435 instances road surface damages (linear, lateral and alligator cracks, potholes, and various types of painting blurs). In order to generate this dataset, we obtained images from several public datasets and augmented it with crowdsourced images, which where manually annotated for further processing. The images were captured under a variety of weather and illumination conditions. In each image, we annotated the bounding box representing the location and type of damage and its extent. Second, we trained different types generic object detection methods, both traditional (an LBP-cascaded classifier) and deep learning-based, specifically, MobileNet and RetinaNet, which are amenable for embedded and mobile and implementations with an acceptable performance for many applications. We compared the accuracy and inference time of all these models with others in the state of the art, achieving higher accuracies in all the eight classes present in the dataset introduced by researchers at the University of Tokyo, and in other related works, with a lower inference time.

show abstract

Section: Related Workmentioning

confidence: 99%

Section: Utilized Datasetmentioning

confidence: 99%

Section: Trainingmentioning

confidence: 99%

See 1 more Smart Citation

Road Damage Detection Acquisition System Based on Deep Neural Networks for Physical Asset Management

Angulo¹,

Vega-Fernández²,

Aguilar-Lobo³

et al. 2019

Lecture Notes in Computer Science

View full text Add to dashboard Cite

show abstract

“…DL methods have been successfully used to object detection [ 20 ] in several applications, such as agriculture and environmental studies [ 21 , 22 ], urban infrastructure [ 23 ] and health analysis [ 24 ]. Thus far, solely few works have been developed to detect manholes using DL ([ 25 ] and [ 26 ]).…”

Section: Introductionmentioning

confidence: 99%

“…Furthermore, one-stage methodologies have lower computational processing costs than two-stage approaches [ 20 , 30 ]. One-stage methods typically use the VGG and ResNet as network backbone [ 31 , 32 ], which have shown good results even compared to the DenseNet backbone [ 23 ]. ResNet backbones (ResNet-50 and ResNet-101) are used to analyze the effect of their depth on the RetinaNet classification model.…”

Section: Introductionmentioning

confidence: 99%

Storm-Drain and Manhole Detection Using the RetinaNet Method

Santos

Marcato

Silva

et al. 2020

Sensors

View full text Add to dashboard Cite

As key-components of the urban-drainage system, storm-drains and manholes are essential to the hydrological modeling of urban basins. Accurately mapping of these objects can help to improve the storm-drain systems for the prevention and mitigation of urban floods. Novel Deep Learning (DL) methods have been proposed to aid the mapping of these urban features. The main aim of this paper is to evaluate the state-of-the-art object detection method RetinaNet to identify storm-drain and manhole in urban areas in street-level RGB images. The experimental assessment was performed using 297 mobile mapping images captured in 2019 in the streets in six regions in Campo Grande city, located in Mato Grosso do Sul state, Brazil. Two configurations of training, validation, and test images were considered. ResNet-50 and ResNet-101 were adopted in the experimental assessment as the two distinct feature extractor networks (i.e., backbones) for the RetinaNet method. The results were compared with the Faster R-CNN method. The results showed a higher detection accuracy when using RetinaNet with ResNet-50. In conclusion, the assessed DL method is adequate to detect storm-drain and manhole from mobile mapping RGB images, outperforming the Faster R-CNN method. The labeled dataset used in this study is available for future research.

show abstract

Study on image diagnosis of timber houses damaged by earthquake using deep learning

Chida

Takahashi

2021

Japan Architectural Review

View full text Add to dashboard Cite

Image processing (semantic segmentation and morphological processing) for the imavge diagnosis of timber houses damaged by earthquake was studied and the following three aspects were revealed.1 Comparing the performance of the models trained with real datasets and chromakeyed datasets, the validation correctness of chromakeyed models achieved comparable or better accuracy than real models (up to 1.38 times more accurate in Frequency Weighted Intersection over Union (FWIoU)). Thus, the usefulness of chromakeyed datasets was confirmed for creating image databases for deep learning. 2 In the image diagnosis results, damage could be assessed with the same accuracy as the investigator if the damage rate was calculated to be 60% or more, unless the damage was unrecognizable due to shielding or resolution problems. This implies the feasibility of image diagnosis for quantitative damage assessment of timber houses. 3 Considering the installation of the estimation methodology of the deformation angle into the image diagnosis, the guideline methodology, and previous experimental values were analyzed. If multiple damage degrees were detected, it could be narrowed down to limited estimation results. This suggests the possibility of installation of estimation methodology into the image diagnosis. Keywords damage assessment, deep learning, image diagnosis, image processing, timber houseThis is an open access article under the terms of the Creative Commons Attribution-NonCommercial-NoDerivs License, which permits use and distribution in any medium, provided the original work is properly cited, the use is non-commercial and no modifications or adaptations are made.

show abstract

Road Damage Detection Using RetinaNet

Cited by 75 publications

References 7 publications

Road Damage Detection Acquisition System Based on Deep Neural Networks for Physical Asset Management

Road Damage Detection Acquisition System Based on Deep Neural Networks for Physical Asset Management

Storm-Drain and Manhole Detection Using the RetinaNet Method

Study on image diagnosis of timber houses damaged by earthquake using deep learning

Contact Info

Product

Resources

About