Tiny CNN for feature point description for document analysis: approach and dataset

作     者：Sheshkus, Alexander Vladimirovich Chirvonaya, Anastasiya Nikolaevna Arlazarov, Vladimir Lvovich 

作者机构：Moscow Institute for Physics Technology Moscow Region Institutskiy per. 9 Dolgoprudny141701 Russia Institute for Systems Analysis Federal Research Center "Computer Science Control" of Russian Academy of Sciences pr. 60-letiya Oktyabrya 9 Moscow117312 Russia Smart Engines Service LLC pr. 60-letiya Oktyabrya 9 Moscow117312 Russia National University of Science Technology "MISIS" Leninskiy prospect 4 Moscow119049 Russia 

出 版 物：《Computer Optics》 (Comput. Opt.)

年 卷 期：2022年第46卷第3期

页      面：429-435页

核心收录：

学科分类：070207[理学-光学] 07[理学] 08[工学] 0835[工学-软件工程] 0714[理学-统计学（可授理学、经济学学位）] 0803[工学-光学工程] 0701[理学-数学] 0812[工学-计算机科学与技术（可授工学、理学学位）] 0702[理学-物理学] 

基　　金：Acknowledgements: This work was supported by the Russian Foundation for Basic ResearchThis work was supported by the Russian Foundation for Basic Research (projects 18-29-26033 and 19-29-09064) 

主　　题：Template matching 

摘      要：In this paper, we study the problem of feature points description in the context of document analysis and template matching. Our study shows that specific training data is required for the task especially if we are to train a lightweight neural network that will be usable on devices with limited computational resources. In this paper, we construct and provide a dataset of photo and syn-thetically generated images and a method of training patches generation from it. We prove the ef-fectiveness of this data by training a lightweight neural network and show how it performs in both general and documents patches matching. The training was done on the provided dataset in comparison with HPatches training dataset and for the testing, we solve HPatches testing framework tasks and template matching task on two publicly available datasets with various documents pic-tured on complex backgrounds: MIDV-500 and MIDV-2019. © 2022, Institution of Russian Academy of Sciences. All rights reserved.