Negative-Supervised Cascaded Deep Learning for Traffic Sign Classification
In this paper, we propose a novel deep learning framework for object classification called negative-supervised cascaded deep learning. There are two hierarchies in our cascaded method: the first one is a convolutional neural network trained on positive-on
- PDF / 1,519,179 Bytes
- 9 Pages / 439.37 x 666.142 pts Page_size
- 13 Downloads / 228 Views
iversity of Chinese Academy of Sciences, Beijing, China 2 Beijing Key Laboratory of IOT Information Security, Institute of Information Engineering, Chinese Academy of Sciences, Beijing, China {xiekaixuan,geshiming,yangrui,luxiang,sunlimin}@iie.ac.cn
Abstract. In this paper, we propose a novel deep learning framework for object classification called negative-supervised cascaded deep learning. There are two hierarchies in our cascaded method: the first one is a convolutional neural network trained on positive-only samples, which is used to select supervisory samples from a negative library. The second one is inherited from the trained first CNN. It is trained on positive and negative samples, which are selected from domain related database by utilizing negative-supervised mechanism. Experiments are applied this idea to traffic sign classification using two classic convolutional neural networks, LeNet-5 and AlexNet as baselines. Classification rates improved by 0.7% and 1.1% with LeNet-5 and AlexNet respectively, which demonstrates the efficiency and superiority of our proposed framework. Keywords: Convolutional neural network · Deep learning supervised · Object classification · Traffic sign classification
1
· Negative-
Introduction
Traffic signs play an important role in our daily life. They define a visual language providing useful information, which makes the driving safe and convenient. In intelligent transportation system (ITS), traffic sign recognition is a critical step for advance driver assistance system (ADAS) and autonomous intelligent vehicles[1]. Traffic sign recognition has two tasks: finding the locations and sizes of traffic signs in natural scene images (traffic sign detection) and classifying the detected traffic signs to their specific sub-classes (traffic sign classification)[2]. In this paper, we focus on traffic sign classification. Due to the complex outdoor environment such as viewpoint variations, bad lighting conditions, motion-blur, occlusions, sun glare, physical damage, colors fading, clustered backgrounds, low resolution and so on, traffic signs are rotated, blurred, damaged and degenerated, which is shown in Fig. 4. Traffic sign recognition is more challenging in comparison with indoor object classification tasks, such as character and face recognition. The uncertainty and ambiguity of traffic signs do not stop the pace of research. To address traffic signs classification, many methods have been proposed according to existing survey literatures[3][4]. c Springer-Verlag Berlin Heidelberg 2015 H. Zha et al. (Eds.): CCCV 2015, Part I, CCIS 546, pp. 249–257, 2015. DOI: 10.1007/978-3-662-48558-3 25
250
K. Xie et al.
Traditional traffic sign classification has been approached with a number of popular machine learning methods, such as support vector machines[5], linear discriminant analysis[6], etc. These methods need to extract of features, such as Histogram of Gradients (HoG)[7], Local Binary Pattern (LBP)[8], Integral Channel Features[9] first. Recently, deep learning such as Convolutional Neural Networks (CNNs), was proposed for
Data Loading...