Learning Spatial Relationships between Samples of Patent Image Shapes
This work addresses a domain-specific problem for patent image analysis, offering an incremental improvement by adapting deep learning techniques to handle variations in binary image generation.
The authors tackled the challenging problem of binary image classification and retrieval for patent documents, which is complicated by variations in image generation mechanisms. They proposed a method using non-Euclidean geometric neural networks to learn spatial relationships, achieving results that outperform existing methods on a patent image dataset benchmark.
Binary image based classification and retrieval of documents of an intellectual nature is a very challenging problem. Variations in the binary image generation mechanisms which are subject to the document artisan designer including drawing style, view-point, inclusion of multiple image components are plausible causes for increasing the complexity of the problem. In this work, we propose a method suitable to binary images which bridges some of the successes of deep learning (DL) to alleviate the problems introduced by the aforementioned variations. The method consists on extracting the shape of interest from the binary image and applying a non-Euclidean geometric neural-net architecture to learn the local and global spatial relationships of the shape. Empirical results show that our method is in some sense invariant to the image generation mechanism variations and achieves results outperforming existing methods in a patent image dataset benchmark.