|
Sketching is an efficient way for recording and conveying ideas, especially in the early stages of design and idea externalizing. Main challenge in this area comes from the ambiguity of sketch. In this paper, we introduce a method of freehand sketchy shape recognition, in which we adapt the relevance feedback for sketchy shapes recognition to capturing user’s intents. Firstly, the candidate sketchy objects are extracted by means of similarity calculation, based on the establishment of feature-based vector model of freehand sketches. Secondly, the relevance feedback are used to capture the users’ input intends and refine the recognition results incrementally by rebuilding the vector and re-weighting the feature-based vector model alone with the user’s relevance judgement. Experiments prove the proposed method both effective and efficient. |
|
Keywords:Sketchy shape recognition, Feature vector, Relevance feedback, Re-weighting, Vector point movement |
|