Sketch recognition

About: Sketch recognition is a research topic. Over the lifetime, 1611 publications have been published within this topic receiving 40284 citations.

Visual Recognition of Sketched Symbols

Abstract: Diagrams are an essential means of capturing and communicating information in many different domains. They are also a valuable part of the early design process, helping us explore ideas and solutions in an informal environment. This paper presents a new approach to sketched symbol recognition that preserves as much of the visual nature of the symbol as possible. Our method is robust to differences in drawing style, computationally efficient, and achieves excellent performance for several different domains. Author Keywords Sketch Recognition, symbol recognition, vision recognition algorithms

HUNCH 1972: A Second Experiment in Sketch Recognition or: ‘I Know the Concept of Your Concept of Interpolation’

Abstract: James Taggart developed the sketch recognition system HUNCH at MIT in 1972. Rather than the user trying to understand the software in order to progress a drawing, in the case of HUNCH, the software observed the user sketching. It enabled a conversation between user and software through the sketch as a medium. HUNCH was one component of the Architecture Machine, created in the Architecture Machine Group (Arch Mac), run by Nicholas Negroponte at MIT between 1967 and 1985 and a brainchild of cross-fertilization between architecture, computer sciences and cybernetics in the early 1970s. One of HUNCHs objectives was to enable even the layman to ‘design’ a dream home. The paper presents HUNCH’s concept as AI-equipped ‘partner’ for the architect, designer or layman.

A neural-network-based sketch recognition system

Abstract: Children really enjoy painting. Painting is not only full of fun but also beneficial for children to further develop their physical skills (e.g., hand eye coordination, fine motor skills, and gross motor skills), express their creativity, boost their self-confidence, etc. On the other hand, storytelling is also a joyful and educational activity for children. In this paper, we develop an interactive platform called An Interactive Painting and Storytelling Platform (AIPSP) on which the two activities of painting and storytelling mingle. Via the proposed AIPSP, children can create a story simply via sketching some main characters of the story on a drawing board. Via these characters and interactions, some creative and funny stories can be created.

A dynamic stroke segmentation technique for sketched symbol recognition

Abstract: In this paper, we address the problem of ink parsing, which tries to identify distinct symbols from a stream of pen strokes. An important task of this process is the segmentation of the users' pen strokes into salient fragments based on geometric features. This process allows users to create a sketch symbol varying the number of pen strokes, obtaining a more natural drawing environment. The proposed sketch recognition technique is an extension of LR parsing techniques, and includes ink segmentation and context disambiguation. During the parsing process, the strokes are incrementally segmented by using a dynamic programming algorithm. The segmentation process is based on templates specified in the productions of the grammar specification from which the parser is automatically constructed.

Forensic sketch recognition using user specific facial region

Abstract: Forensic sketches play an important role in criminal identification process. These sketches are drawn by forensic artists on the basis of the description provided by an eyewitness or victim. These sketches are publicised to get some clues to reveal the identity of the criminals. A faster way to identify the criminals is to match the forensic sketch with some government agency mug-shot database. In the process of drawing a sketch, the description provided by the eyewitness often includes some unique facial details comprising the deviations from an average face. For example, some spot on face, mole, scar, cuts, etc. Study says that, the information which is more uncommon is more likely to be last in the memory. In this paper, we suggest a sketch to mug-shot matching approach called difference vector-based matching DVBM, which utilises deviations present in facial regions to measure the similarity between a sketch and a mug-shot image. The method is tested over a dataset containing 112 sketches and a large mug-shot gallery of 7,112 images. The results generated using DVBM are compared with standard face matcher and show considerable improvement in matching accuracy.