联合语义部件的鸟类图像细粒度识别-计算机辅助设计与图形学学报.PDF

30 8 Vol.30 No.8 2018 8 Journal of Computer-Aided Design Computer Graphics Aug. 2018 联合语义部件的鸟类图像细粒度识别 1,2) 1)* 赵毅力 , 徐 丹 1) (云南大学信息学院 昆明 650091) 2) (西南林业大学大数据与智能工程学院 昆明 650224) (danxu@) 摘 要: 由于子类别的高度相似性引起的类间微小差异, 以及姿态、尺度和旋转方面的类内变化, 使得细粒度图像识 别成为一个具有挑战性的计算机视觉问题. 为了对鸟类图像进行细粒度识别, 提出一种联合语义部件的深度卷积神 经网络模型. 该模型由2 个子网络组成: 一个是语义部件检测子网, 使用深度残差网络对鸟类图像语义部件进行精确 定位; 另一个是分类子网, 使用三路深度残差网络对检测子网检测到的语义部件进行联合分类. 收集了一个新的鸟 类图像数据集YUB-200-2017, 用于鸟类图像细粒度识别实验. 结果表明, 在YUB-200-2017 和CUB-200-2011 数据集 上, 文中方法具有较高的语义部件检测精度和识别准确率. 关键词: 细粒度图像识别; 语义部件检测; 深度学习; 卷积神经网络 中图法分类号: TP391.41 DOI: 10.3724/SP.J.1089.2018.16781 Joint Semantic Parts for Fine-Grained Bird Images Recognition Zhao Yili1,2) and Xu Dan1)* 1) (School of Information, Yunnan University, Kunming 650091) 2) (School of Big Data and Intelligent Engineering, Southwest Forestry University, Kunming 650224) Abstract: Fine-grained image recognition is a challenging computer vision problem, due to small inter-class variations caused by highly similar subordinate categories, and the large intra-class variations in poses, scales and rotations. In order to perform fine-grained recognition on bird images, this paper proposes a deep convolution neural networks model collaborated with semantic parts detection. The model consists of two modules, one module is a parts detector network, and another module is a three-stream classification network based on deep residual network. In the meantime, a new bird images dataset was collected and labeled to fa- cility the research of fine-grained bird images recognition. Experiment results on YUB-200-2017 and CUB-200-2011 i