Environment Estimation Based on Inter-object Relations

Abstract

Here we presented the model for environment estimation based on the interobject relations. The uniqueness of this model is that it relies on the results of image segmentation and predicts environment based on the object properties and their relationships. The model provides context by describing the environment where particular objects’ setup can occur. Many other works reported usefulness of contextual information for image processing problems. In particular, [16] had proven that contextual information can improve the accuracy of the verification process up to 16%. As part of the feature work we plan to assess the improvement that our model brings to the overall image segmentation process as part of the Algorithm selection platform.