Image annotation of power grid objects based on convolutional neural networks
Luo Wang, Min Feng, Qiang Fan, Qiwei Peng, Guozhi Li, Xiaolong Hao, Lei Yu, Xia Yuan · 2016
In this paper, we propose a novel method to annotate the image of power grid objects (i.e., the electric equipment, the workers with different behaviors). This method is based on the convolutional neural networks (CNN). First, we obtain the attribute list of the image under the multi-label networks. Second, we employ the attribute-specific segmentation model to annotate the image. In this paper, we build an image database for power grid objects which consists of a large number of images, such as the electric equipment and the workers with different behavior. The experimental results demonstrate the good performance of the proposed method.