发明名称 |
Font recognition and font similarity learning using a deep neural network |
摘要 |
A convolutional neural network (CNN) is trained for font recognition and font similarity learning. In a training phase, text images with font labels are synthesized by introducing variances to minimize the gap between the training images and real-world text images. Training images are generated and input into the CNN. The output is fed into an N-way softmax function dependent on the number of fonts the CNN is being trained on, producing a distribution of classified text images over N class labels. In a testing phase, each test image is normalized in height and squeezed in aspect ratio resulting in a plurality of test patches. The CNN averages the probabilities of each test patch belonging to a set of fonts to obtain a classification. Feature representations may be extracted and utilized to define font similarity between fonts, which may be utilized in font suggestion, font browsing, or font recognition applications. |
申请公布号 |
US9501724(B1) |
申请公布日期 |
2016.11.22 |
申请号 |
US201514734466 |
申请日期 |
2015.06.09 |
申请人 |
ADOBE SYSTEMS INCORPORATED |
发明人 |
Yang Jianchao;Wang Zhangyang;Brandt Jonathan;Jin Hailin;Shechtman Elya;Agarwala Aseem Omprakash |
分类号 |
G06K9/00;G06K9/36;G06K9/66;G06K9/62;G06K9/68;G06T3/40 |
主分类号 |
G06K9/00 |
代理机构 |
Shook, Hardy & Bacon L.L.P. |
代理人 |
Shook, Hardy & Bacon L.L.P. |
主权项 |
1. A non-transitory computer storage medium comprising computer-useable instructions that, when used by one or more computing devices, cause the one or more computing devices to perform operations comprising:
receiving one or more text images, each of the one or more text images including a corresponding font label; synthesizing the one or more text images to introduce variances in the one or more text images; generating one or more training images that include the variances; cropping the one or more training images into training patches that are utilized as an input to a convolutional neural network (CNN); training the CNN with the training patches, the CNN comprising a plurality of convolutional layers and a plurality of fully connected layers; and producing a distribution of classified text images. |
地址 |
San Jose CA US |