Multi-Task Vehicle Detection With Region-of-Interest Voting

Chu, Wenqing; Liu, Yao; Shen, Chen; Cai, Deng; Hua, Xian-Sheng

doi:10.1109/tip.2017.2762591

Cited by 82 publications

(58 citation statements)

References 49 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Comparing to YOLO, YOLOv2 achieves higher accuracy and higher speed. To better handle the detection problem of vehicles in complex conditions, Chu et al [21] proposed a vehicle detection scheme based on multitask deep CNN in which learning is trained on four tasks: category classification, bounding box regression, overlap prediction, and subcategory classification. A region of interest voting scheme and multilevel localization are then used to further improve detection accuracy and reliability.…”

Section: Theoretical Basismentioning

confidence: 99%

Improving Faster R‐CNN Framework for Fast Vehicle Detection

Nguyen

2019

Mathematical Problems in Engineering

View full text Add to dashboard Cite

Vision-based vehicle detection plays an important role in intelligent transportation systems. With the fast development of deep convolutional neural networks (CNNs), vision-based vehicle detection approaches have achieved significant improvements compared to traditional approaches. However, due to large vehicle scale variation, heavy occlusion, or truncation of the vehicle in an image, recent deep CNN-based object detectors still showed a limited performance. This paper proposes an improved framework based on Faster R-CNN for fast vehicle detection. Firstly, MobileNet architecture is adopted to build the base convolution layer in Faster R-CNN. Then, NMS algorithm after the region proposal network in the original Faster R-CNN is replaced by the soft-NMS algorithm to solve the issue of duplicate proposals. Next, context-aware RoI pooling layer is adopted to adjust the proposals to the specified size without sacrificing important contextual information. Finally, the structure of depthwise separable convolution in MobileNet architecture is adopted to build the classifier at the final stage of the Faster R-CNN framework to classify proposals and adjust the bounding box for each of the detected vehicle. Experimental results on the KITTI vehicle dataset and LSVH dataset show that the proposed approach achieved better performance compared to original Faster R-CNN in both detection accuracy and inference time. More specific, the performance of the proposed method is improved comparing with the original Faster R-CNN framework by 4% on the KITTI test set and 24.5% on the LSVH test set.

show abstract

Section: Theoretical Basismentioning

confidence: 99%

Improving Faster R‐CNN Framework for Fast Vehicle Detection

Nguyen

2019

Mathematical Problems in Engineering

View full text Add to dashboard Cite

show abstract

“…The proposed approach uses both the feature values along with the class labels to generate classification locations. Chu et al [8] developed a deep convolutional neural network with region of interest voting. Offset direction of each ROI boundary is predicted by the proposed CNN model.…”

Section: Related Workmentioning

confidence: 99%

Vehicle Detection from Images using Deep Fully Convolutional Networks

Bhargavi¹,

Rao²

2019

IJRTE

View full text Add to dashboard Cite

Vehicle detection provides the facilitation of traffic planning and management. It also helps in finding suspected and stolen vehicles. Although it has many applications, it is a very complex problem due to variations in vehicle type and size. As the amount of vehicle types are very high the models find it hard to classify the correct type of the vehicle. In this paper, we are proposing a vehicle detection model based on YoloV3 convolutional neural network architecture with custom backbone. Our proposed backbone in the Yolov3 architecture helps classify the different types of vehicles accurately. This makes the classification of the images at pixel level and predicts the regression based ROI bounding box for the classified vehicles in the images. The model contains features extracted at different kernel sizes to find the features at multiple scales which will then be concatenated. Experiments were performed on the Kitti vehicle detection dataset have shown the superior performance of our proposed model.

show abstract

“…Chu proposes a novel vehicle detection scheme based on multi-task deep Convolutional Neural Networks (CNNs) and Region-Of-Interest (RoI) voting. In the design of the CNN architecture, this study enriches the supervised information with subcategory, region overlap, bounding-box regression, and a category for each training RoI as a multi-task learning framework [29].…”

Section: Fault Diagnosis Of a Vehiclementioning

confidence: 99%

An Integrated Self-Diagnosis System for an Autonomous Vehicle Based on an IoT Gateway and Deep Learning

et al. 2018

View full text Add to dashboard Cite

This paper proposes "An Integrated Self-diagnosis System (ISS) for an Autonomous Vehicle based on an Internet of Things (IoT) Gateway and Deep Learning" that collects information from the sensors of an autonomous vehicle, diagnoses itself, and the influence between its parts by using Deep Learning and informs the driver of the result. The ISS consists of three modules. The first In-Vehicle Gateway Module (In-VGM) collects the data from the in-vehicle sensors, consisting of media data like a black box, driving radar, and the control messages of the vehicle, and transfers each of the data collected through each Controller Area Network (CAN), FlexRay, and Media Oriented Systems Transport (MOST) protocols to the on-board diagnostics (OBD) or the actuators. The data collected from the in-vehicle sensors is transferred to the CAN or FlexRay protocol and the media data collected while driving is transferred to the MOST protocol. Various types of messages transferred are transformed into a destination protocol message type. The second Optimized Deep Learning Module (ODLM) creates the Training Dataset on the basis of the data collected from the in-vehicle sensors and reasons the risk of the vehicle parts and consumables and the risk of the other parts influenced by a defective part. It diagnoses the vehicle's total condition risk. The third Data Processing Module (DPM) is based on Edge Computing and has an Edge Computing based Self-diagnosis Service (ECSS) to improve the self-diagnosis speed and reduce the system overhead, while a V2X based Accident Notification Service (VANS) informs the adjacent vehicles and infrastructures of the self-diagnosis result analyzed by the OBD. This paper improves upon the simultaneous message transmission efficiency through the In-VGM by 15.25% and diminishes the learning error rate of a Neural Network algorithm through the ODLM by about 5.5%. Therefore, in addition, by transferring the self-diagnosis information and by managing the time to replace the car parts of an autonomous driving vehicle safely, this reduces loss of life and overall cost.

show abstract

Multi-Task Vehicle Detection With Region-of-Interest Voting

Cited by 82 publications

References 49 publications

Improving Faster R‐CNN Framework for Fast Vehicle Detection

Improving Faster R‐CNN Framework for Fast Vehicle Detection

Vehicle Detection from Images using Deep Fully Convolutional Networks

An Integrated Self-Diagnosis System for an Autonomous Vehicle Based on an IoT Gateway and Deep Learning

Contact Info

Product

Resources

About