Improved Faster R-CNN Traffic Sign Detection Based on a Second Region of Interest and Highly Possible Regions Proposal Network

Abstract

Traffic sign detection systems provide important road control information for unmanned driving systems or auxiliary driving. In this paper, the Faster region with a convolutional neural network (R-CNN) for traffic sign detection in real traffic situations has been systematically improved. First, a first step region proposal algorithm based on simplified Gabor wavelets (SGWs) and maximally stable extremal regions (MSERs) is proposed. In this way, the region proposal a priori information is obtained and will be used for improving the Faster R-CNN. This part of our method is named as the highly possible regions proposal network (HP-RPN). Second, in order to solve the problem that the Faster R-CNN cannot effectively detect small targets, a method that combines the features of the third, fourth, and fifth layers of VGG16 to enrich the features of small targets is proposed. Third, the secondary region of interest method to enhance the feature of detection objects and improve the classification capability of the Faster R-CNN is proposed. Finally, a method of merging the German traffic sign detection benchmark (GTSDB) and Chinese traffic sign dataset (CTSD) databases into one larger database to increase the number of database samples is proposed. Experimental results show that our method improves the detection performance, especially for small targets.

Robust In-Plane Structures Oscillation Monitoring by Terrestrial Photogrammetry

“…Network evaluation is performed applying mean average precision (mAP) open source code [34], by using 138 images, including 139 Targets and 138 Track-points. Average precision (AP), mean average precision (mAP), recall, and intersection over union (IoU) [35,36] are used for the network's performance evaluation, where IoU refers to the degree of coincidence between the detected area and the ground truth area as:…”

The Development of a Skin Cancer Classification System for Pigmented Skin Lesions Using Deep Learning

“…In the present study, we used the faster, region-based convolutional neural network (Faster R-CNN, or FRCNN) algorithm, which is a result of merging region proposal network (RPN) and Fast R-CNN algorithms, into a single network [ 13 , 14 ]. The pioneering work of region-based target detection began with the region-based convolutional neural network (R-CNN), including three modules: regional proposal, vector transformation, and classification [ 15 , 16 ]. Spatial pyramid pooling (SPP)-net optimized the R-CNN and improved detection performance [ 16 , 17 ].…”

“…The pioneering work of region-based target detection began with the region-based convolutional neural network (R-CNN), including three modules: regional proposal, vector transformation, and classification [ 15 , 16 ]. Spatial pyramid pooling (SPP)-net optimized the R-CNN and improved detection performance [ 16 , 17 ]. Fast R-CNN combines the essence of SPP-net and R-CNN, and introduces a multi-task loss function, which is what makes the training and testing of the whole network so functional [ 16 , 18 ].…”

“…Spatial pyramid pooling (SPP)-net optimized the R-CNN and improved detection performance [ 16 , 17 ]. Fast R-CNN combines the essence of SPP-net and R-CNN, and introduces a multi-task loss function, which is what makes the training and testing of the whole network so functional [ 16 , 18 ]. FRCNN merges RPN and Fast R-CNN into a unified network by sharing the convolutional features with “attention” mechanisms, which greatly improves both the time and accuracy of target detection [ 13 , 16 ].…”

“…Fast R-CNN combines the essence of SPP-net and R-CNN, and introduces a multi-task loss function, which is what makes the training and testing of the whole network so functional [ 16 , 18 ]. FRCNN merges RPN and Fast R-CNN into a unified network by sharing the convolutional features with “attention” mechanisms, which greatly improves both the time and accuracy of target detection [ 13 , 16 ]. Indeed, FRCNN has shown higher detection performance in the biomedical filed than other state-of-the-art methods, such as support vector machines (SVMs), visual geometry Group-16 (VGG-16), single shot multibox detectors (SSDs), and you only look once (YOLO), in terms of time and accuracy [ 19 , 20 , 21 ].…”