A closer look at Faster R-CNN for vehicle detection

Fan, Quanfu; Brown, Lisa M.; Smith, J. R.

doi:10.1109/ivs.2016.7535375

Cited by 175 publications

(100 citation statements)

References 15 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Inputs to Faster R‐CNN can be scaled to decrease the inappropriate effects scale variations in objects (Ren et al., ; Fan et al., ), although we do not apply any scaling on our images, which are taken from an approximately constant distance between the damages and camera and have insignificant scale invariances. Ren et al.…”

Section: Database and Implementation Detailsmentioning

confidence: 99%

“…For the last step, Fast R-CNN takes the object proposals generated in step three and is trained with the initial parameters trained in step three. As RPN may produce more than 2,000 object proposals for an image, which causes costly computations and may decrease the accuracy of object detection (Girshick, 2015;Ren et al, 2016;Fan et al, 2016), outputs of RPN are sorted based on the score of its softmax layer, and the first 2,000 objects proposals (if there are more than 2,000 generated proposals) with the highest scores are fed into Fast R-CNN in the second stage of training. Using the methods of Ren et al (2016) for the fourth stage of training as well as the testing stage, the first 300 object proposals with the highest scores are used to increase the speed of detection.…”

Section: Faster R-cnn By Sharing Cnn Between Rpn and Fast R-cnnmentioning

confidence: 99%

“…Ren et al (2016) acknowledged that in their study, they did not choose initial parameters of Faster R-CNN carefully. Although Fan et al (2016) studied the influence of these parameters on the performance of Faster R-CNN, they did not introduce a straightforward way to choose these parameters, so we had to find them via trial-and-error. To find the optimal anchor sizes and ratios, we examined 27 combinations of ratios from nine different ratios (0.2, 0.35, 0.5, 0.85, 1, 1.15, 1.7, 1.85, and 2) and two combinations of sizes from six different anchor sizes (96, 128, 192, 256, 384, and 512).…”

Section: Implementation Detailsmentioning

confidence: 99%

“…Inputs to Faster R-CNN can be scaled to decrease the inappropriate effects scale variations in objects Fan et al, 2016), although we do not apply any scaling on our images, which are taken from an approximately constant distance between the damages and camera and have insignificant scale invariances. Ren et al (2016) applied scaling of 600 on the shorter side of images (with the average resolution of 500 × 375), and slid the sliding CONV layer of RPN with stride 16 on feature map.…”

Section: Implementation Detailsmentioning

confidence: 99%

See 3 more Smart Citations

Autonomous Structural Visual Inspection Using Region‐Based Deep Learning for Detecting Multiple Damage Types

Cha

Choi

Suh

et al. 2017

Computer aided Civil Eng

1,127

515

View full text Add to dashboard Cite

Computer vision‐based techniques were developed to overcome the limitations of visual inspection by trained human resources and to detect structural damage in images remotely, but most methods detect only specific types of damage, such as concrete or steel cracks. To provide quasi real‐time simultaneous detection of multiple types of damages, a Faster Region‐based Convolutional Neural Network (Faster R‐CNN)‐based structural visual inspection method is proposed. To realize this, a database including 2,366 images (with 500 × 375 pixels) labeled for five types of damages—concrete crack, steel corrosion with two levels (medium and high), bolt corrosion, and steel delamination—is developed. Then, the architecture of the Faster R‐CNN is modified, trained, validated, and tested using this database. Results show 90.6%, 83.4%, 82.1%, 98.1%, and 84.7% average precision (AP) ratings for the five damage types, respectively, with a mean AP of 87.8%. The robustness of the trained Faster R‐CNN is evaluated and demonstrated using 11 new 6,000 × 4,000‐pixel images taken of different structures. Its performance is also compared to that of the traditional CNN‐based method. Considering that the proposed method provides a remarkably fast test speed (0.03 seconds per image with 500 × 375 resolution), a framework for quasi real‐time damage detection on video using the trained networks is developed.

show abstract

Section: Database and Implementation Detailsmentioning

confidence: 99%

Section: Faster R-cnn By Sharing Cnn Between Rpn and Fast R-cnnmentioning

confidence: 99%

Section: Implementation Detailsmentioning

confidence: 99%

Section: Implementation Detailsmentioning

confidence: 99%

See 2 more Smart Citations

Autonomous Structural Visual Inspection Using Region‐Based Deep Learning for Detecting Multiple Damage Types

Cha

Choi

Suh

et al. 2017

Computer aided Civil Eng

1,127

515

View full text Add to dashboard Cite

show abstract

“…To obtain real time implementations, pre-training schemes [27] are used even with low resolution images [28]. Based on object proposal algorithms, two stage CNN models integrate region proposal and classification in a single architecture, such as Fast R-CNN [29] and faster R-CNN [30] based models for vehicle detection and classification [31][32] [33]. Motivated by safety measures, helmet detection in motorcycle riders has inspired research using geometrical features [34], hand crafted features (HOG, SIFT, LBP, CHT [35] [36] [12]), neuro-fuzzy detectors [37] and neural networks [38].…”

Section: Introductionmentioning

confidence: 99%

Motorcycle detection and classification in urban Scenarios using a model based on Faster R-CNN

Espinosa

Velastín

Branch

2018

9th International Conference on Pattern Recognition Systems (ICPRS 2018)

View full text Add to dashboard Cite

This paper introduces a Deep Learning Convolutional NeutralNetwork model based on Faster-RCNN for motorcycle detection and classification on urban environments. The model is evaluated in occluded scenarios where more than 60% of the vehicles present a degree of occlusion. For training and evaluation, we introduce a new dataset of 7500 annotated images, captured under real traffic scenes, using a drone mounted camera. Several tests were carried out to design the network, achieving promising results of 75% in average precision (AP), even with the high number of occluded motorbikes, the low angle of capture and the moving camera. The model is also evaluated on low occlusions datasets, reaching results of up to 92% in AP.

show abstract

Vision‐based vehicle detection for road traffic congestion classification

Chetouane

Mabrouk

Jemili

et al. 2020

Concurrency and Computation

View full text Add to dashboard Cite

SummaryDue to the increasing number of vehicles in circulation in different urban cities, several automatic traffic monitoring systems have been developed. In particular, traffic monitoring systems using roadside cameras are becoming extensively deployed, as they offer imperative technological advantages compared with other traffic monitoring systems. Vehicle detection and traffic congestion classification are two main steps for video‐based traffic congestion detection systems; the associated methods have a deep impact on the performance of the whole system. In this paper, we investigate four selected vehicle detection methods namely Gaussian Mixture Model (GMM), GMM‐Kalman filter, Optical Flow, and ACF object detector in two contexts: urban and highway. Three traffic congestion classification methods are also studied. The comparative study of the different methods allows us to choose the most appropriate ones to be integrated in the framework proposed to solve the traffic issues in the bridge of Bizerte.

show abstract

A closer look at Faster R-CNN for vehicle detection

Cited by 175 publications

References 15 publications

Autonomous Structural Visual Inspection Using Region‐Based Deep Learning for Detecting Multiple Damage Types

Autonomous Structural Visual Inspection Using Region‐Based Deep Learning for Detecting Multiple Damage Types

Motorcycle detection and classification in urban Scenarios using a model based on Faster R-CNN

Vision‐based vehicle detection for road traffic congestion classification

Contact Info

Product

Resources

About