Floorplan analysis has been an active area of research for Document Image Analysis and Recognition (DIAR) and Graphics RECognition (GREC) communities. Over last couple of decades, a lot of research has been carried out for different problems related to floorplan analysis (symbol recognition, symbol localization, symbol spotting, object detection, object recognition). Recent trend of Neural Networks based methods has also attracted the attention of researchers working on floorplan images. The proposed competition will try to attract the attention of all these researchers who were/are involved in such research and proposes the challenge of “Object Detection and Recognition in Floorplan images” for this edition. A new huge labelled dataset will be developed and made available to the research community afterwards, in order to provide a continuously growing common dataset for benchmarking the different works.

Keywords: Floorplan, Object Detection, Object Recognition, Symbol Detection, Symbol Recognitio, Graphics Recognition, Document Image Analysis and understanding, Pattern Recognition, Computer Vision.

Floorplan analysis has been an active area of research for the Document Image Analysis and Recognition (DIAR) community in general and the Graphics RECognition (GREC) community in particular. During last couple of decades ​several works have been published on the problem of symbol recognition, symbol localization, symbol spotting, sketch recognition etc. In past the GREC community has organized many competitions on ​these problems and other problems related to floorplan analysis. Competition on recognition of dashed lines (GREC1995, GREC1997, GREC1999), detection of arcs (GREC2001, GREC2003), symbol recognition (ICPR2000, GREC2003, GREC2005, GREC2007, GREC2013). A lot of investigation has been carried out for producing labelled datasets for the above stated problems related to the floorplan analysis.

In the past due to lack of publicly available labelled datasets, very few methods dealt floorplan analysis as problem of machine learning. After a few years of less activity around the above stated problems on the floorplan images, an increase of interest has been noted recently … thanks to the Neural Networks based approaches that have gained much attention in the last few years. We feel that this is the right time to work on a labelled dataset for providing a common benchmark to the community. This will motivate the researchers to report results on a common public dataset in addition to home-built datasets for floorplan analysis.

The proposed competition is of great interest to the ICDAR community, as it will allow the researchers to benchmark their methods and algorithms on a huge dataset with real challenging scenarios and difficulties. This will provide an opportunity to the researchers who have worked in past on floorplan analysis to see how their classical approaches perform on the new dataset and in comparison, to the modern approaches of floorplan analysis.

In addition to the ICDAR 2019 competition session, we are in touch with the GREC workshop organizers for presenting our competition, dataset and participant’s methods during upcoming GREC workshop. This will permit us to introduce our dataset and will be interesting to create/boost some activity in the GREC community. This year we are proposing a first competition on floorplan analysis and our goal is to create a series of competitions on floorplan analysis by proposing new tasks in the future editions of ICDAR as well; for continue contributing new challenging datasets to the community.