期刊名称：International Journal of Computer Science and Network Security
印刷版ISSN：1738-7906
出版年度：2020
卷号：20
期号：6
页码：216-222
出版社：International Journal of Computer Science and Network Security
摘要：Deep learning approaches showed significant performance in current computer vision tasks particularly on image classification and object detection. Object detection is growing its popularity for video surveillance systems and recognizing objects. Convolutional neural networks (CNNs) opens a wide path for computer vision applications. In this paper, we proposed a method to enable select target classes for detection, produce an initial detection representation by selecting a specific portion of the image and maintain the trainset for detection model. The method achieved noticeable detection results to detect multi-class objects. However, object detection frameworks face difficulties to localize small objects. The main cause of this problem is to adopt exact feature mapping and extracting strategy. Due to the low pixel value feature extractors unable to map and localize the small objects. To eliminate the issue we designed a convolutional network named Adaptive Feature Scaler (AFS) Convolutional network. The network constructed to localize and extract exact feature data to detect multi-class objects.
