SketchCleanNet — A deep learning approach to the enhancement and correction of query sketches for a 3D CAD model retrieval system

Search and retrieval remains a major research topic in several domains, including computer graphics, vision, engineering design, etc. A search engine requires primarily an input query database of items to from. In engineering, which is the primary context this paper, consists 3D CAD models, such as washers, pistons, connecting rods, from user typically form sketch, attempts capture details model. However, sketches have certain typical defects gaps, over-drawn portions (multi-strokes), Since retrieved results are only good query, need cleaning-up enhancement for better results. deep learning approach proposed improve or clean sketches. Initially, various categories analysed order understand many possible that may occur. dataset cleaned-up enhanced then created based on understanding these defects. Consequently, end-to-end training neural network carried out provide mapping between defective This takes sketch generates sketch. Qualitative quantitative comparisons with other state-of-the-art techniques show effective. The reported using both sketches, it shown developed yields improved