面向AI算法的外观专利图像数据集

作为一种工业产权的外观专利在产业振兴中起着重要作用.外观专利采用的洛迦诺分类标准不支持图像分类,而AI算法直接在洛迦诺分类标准图像数据集并不能有效提升分类精度.因此,对外观专利审查员来说,外观专利的分类检索具有相当大的挑战.为此,本文提出先领域、再功能、后视觉的四级外观专利图像分类新标准,在此分类标准基础上构建了Pat...

Design Patent Image Dataset for AI Algorithm

As one of industrial property rights,the design patents play important roles in industrial revitalization.The Locarno classification standard of design patents does not support image classification.The application of AI technology directly to the Locarno classification image dataset does not improve the classification accuracy.For design patent examiners,the classification and retrieval of the design patents is a huge challenge.Therefore,a new 4-level classification standard which includes domain,function,and vision feature is proposed for design patent images.The Patent-MNIST,which based on the proposed standard,is constructed to test and verify the classification standard.Some classic AI algorithms are implemented on the dataset.The results show that the classification accuracy of classic algorithms such as VGG19 in the PatentMNIST dataset is significantly better than that of using the Locarno classification standard dataset directly.Also,there is a need to further improve performance of the algorithm based on design patents.