The CBDAR international workshop series has a special focus on camera captured documents and textual content in general. The situation surrounding the CBDAR field has been evolving rapidly. The mobile camera industry is expected to expand reaching US$18.8 billion market value by 2021 (source: “CMOS Industry Overview”, Yole, June 2016). We are in the middle of a second explosive growth of mobile camera technologies, driven by new applications such as virtual/augmented reality, drones and robots, that will push the specifications even higher. Dedicated consumer and professional camera-based document scanning solutions are currently in the market. In a 2012 Xerox white paper, on mobile document scanner (“Mobile Document Capture: Scanner vs. Phone Camera”, John Capurso, Vice President of Marketing (Visioneer Inc., a Xerox® Trademark Licensee), 2012), the author concluded that scanning with mobile cameras was a pretty bad idea, arguing that dedicated devices should be used for functions such as scanning summarising the conclusion as follows: “Would you use your phone’s camera to shoot your vacation to Europe? Would you use your phone’s GPS as your main GPS device in your car?”. It can be easily argued that this is not how the future played out, and document understanding based on mobile phones is already becoming a popular alternative to scanners in specific situations. There are a multitude of Apps for document scanning that offer advanced functionality such as automatically detecting and cropping pages and automatic image enhancement.
At the same time, text recognition in the wild has improved substantially over the past years, and is integrated in translation services, while APIs are available that permit integrating text detection and recognition functionality in any App – for example OpenCV’s Scene Text Detection3 and Recognition4 modules or Google’s Text Recognition API5.
We therefore consider the organisation of CBDAR 2021, providing a forum to explore and discuss new research directions in this area periodically, as timely as ever. CBDAR traditionally attracts researchers from the mainstream computer vision area, and receives about half of the paper submissions from industry. It is therefore a relevant event for ICDAR, both in terms of expanding visibility of the DIAR field but also as an attractor for the industry.