Generation of synthetic documents for performance evaluation of symbol recognition & spotting systems

Abstract

In this paper we present a robust system of symbol recognition using a structural approach. Our key objective here is to provide a system, equaling the statistical ones in robustness concerning the recognition, to apply next to localization. To do it we have investigated two particular structural methods: the straight line detection using Hough Transform and the vector templates matching. Experiments done on the GREC2003 database show how their combination allows to obtain high recognition results.

On the Scalability of Graphic Symbol Recognition

“…With just one image by class for the training, it's hard to define the setting of recognition system in order to determine if two images with just some different pixels (see figure 2) represent two different symbols or the same symbol exposed to noise. In this case, a solution is to use a generic degradation model proposed by Kanungo et al [8] or a model dedicated to a particular domain [9], [10], [11] to generate training data. This strategy only works if the model perfectly match with the degradation could be affected a symbol but on large scale a model cannot generate all type of degradation.…”

DocCreator: A New Software for Creating Synthetic Ground-Truthed Document Images

“…Baird et al [27] no n/a n/a n/a no 1990 Zhao et al [28] no n/a n/a n/a no 2005 Delalandre et al [12] no n/a n/a n/a no 2010 Yin et al [30] no n/a n/a n/a no 2013 Mas et al [24] no n/a n/a n/a yes 2016 Seuret et al [35] no yes n/a n/a no 2015…”

“…to automatically create multiple documents with varied contents (in terms of font, background, layout). Alternative approaches consist of re-arranging, in a new way, elements extracted from real images so as to generate (manually, semi-automatically or automatically) multiple semi-synthetic document images [12,30]. Recently, in particular with the advent of deep learning techniques which require huge masses of training data, the need for synthetic data generation seems to be ever-growing.…”

“…; for handwritten documents IAM database [6], RIMES [7], GERMANA [8], etc. ; for graphical documents: chemical symbol database [9], logo databases [10,11], architectural symbol database [12] or musical symbol database CVC-MUSICMA [13]; camera-based document image analysis [14,15]. The International Association for Pattern Recognition, for instance, gathered some interesting datasets [16] mostly used for different conference competitions over the last two decades.…”

Deep neural networks for record counting in historical handwritten documents

“…In the field of document image analysis and recognition these approaches to synthetic pattern generation have been adopted, for instance, to model the process of character degradation [11] or in the field of graphical documents for generating synthetic documents for performance evaluation of symbol recognition systems [12]. In the area of handwriting recognition, cursive fonts have been used to synthetically generate handwritten documents [13].…”