English (English)

Analyzing images

You need to analyze the images in the batch in order to find objects which can form the backbone of your FlexiLayout, i.e. you need to find objects which are present on all of the images and which can be reliably detected by the program.

Another important decision you need to make when analyzing the images is in which order to describe the objects, as this order determines the order in which the program will use the elements when matching the FlexiLayout with images.

The general scenario is as follows:

  1. Look through the images in the batch.
  2. Visually try to find objects which are present on most of the images and occur on each image only once. Such objects will be created in the beginning of FlexiLayout creation.
  3. Find an object (or objects) which can be used as the identifier. This can be the title or any other object which unambiguously identifies the document as belonging to a particular type. The identifier object must be present on all of the documents and, if possible, recognized without errors. We recommend marking the element which will describe the identifier object as required element. If the program fails to find the object corresponding to this element, FlexiLayout matching will be stopped, which will mean that the document does not belong to the document type described by the FlexiLayout
  4. Consider the order in which you will create elements and how they will be arranged in the FlexiLayout tree. The order of elements in the tree will determine the order in which the program will look for the corresponding objects on the images.
  5. Consider the methods that can be used to find each of the objects. In the FlexiLayout, these methods will be described by means of element properties. Start with the most reliable objects, i.e. those which are present on most of the images and which do not occur too often on the same image. Later they can be used as starting points, or reference elements, to look for less reliable elements.
    Be sure to select the strictest search criteria, so that the program formulates the most reliable hypothesis.
    To prevent an uncontrollable growth of the tree of hypotheses, we do not recommend changing the default value in the Number of surviving hypotheses field.
  6. Consider grouping some of the elements into Group elements. Group elements are convenient for testing parts of the FlexiLayout which are independent of one another. For example, a FlexiLayout made up of a total of 100 elements, may include only 3 Group elements at the highest level - the title, the body, and the bottom of the document. Each of the three Group elements may be made up of Group elements which describe smaller portions of the document. Such nesting of Group elements reduces the number of possible search combinations and makes debugging the FlexiLayout easier, as you can work with each part of the FlexiLayout independently.

12.04.2024 18:16:02

Please leave your feedback about this article

Usage of Cookies. In order to optimize the website functionality and improve your online experience ABBYY uses cookies. You agree to the usage of cookies when you continue using this site. Further details can be found in our Privacy Notice.