[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-seo-664823-105":3,"detail-sidebar-cat-0-en-105":79,"doc-detail-664823-en":129},{"code":4,"msg":5,"data":6},0,"ok",{"site_id":7,"language":8,"slug":9,"title":10,"keywords":11,"description":12,"schema_data":13,"social_meta":72,"head_meta":74,"extra_data":76,"updated_unix":78},105,"en","a-neural-representation-of-sketch-drawings","A Neural Representation of Sketch Drawings","","Sketch-RNN is a recurrent neural network that generates stroke-based drawings of common objects. The model is trained on thousands of crude, human-drawn sketch images spanning hundreds of classes. The work introduces a framework for both conditional and unconditional sketch generation in a vector (stroke sequence) format. It also presents robust training methods to produce coherent sketches, and examines the latent space to improve generation quality and enable vector-based drawing generalization.",{"@graph":14,"@context":71},[15,34,54],{"@type":16,"itemListElement":17},"BreadcrumbList",[18,23,27,31],{"item":19,"name":20,"@type":21,"position":22},"https://docshare.wps.com","Home","ListItem",1,{"item":24,"name":25,"@type":21,"position":26},"https://docshare.wps.com/document/","Document",2,{"item":28,"name":29,"@type":21,"position":30},"https://docshare.wps.com/document/research-report/","Research & Report",3,{"item":32,"name":10,"@type":21,"position":33},"https://docshare.wps.com/document/a-neural-representation-of-sketch-drawings/664823/",4,{"url":32,"name":10,"@type":35,"image":36,"author":41,"headline":10,"publisher":44,"fileFormat":47,"inLanguage":8,"description":12,"dateModified":48,"datePublished":48,"encodingFormat":47,"isAccessibleForFree":49,"interactionStatistic":50},"DigitalDocument",{"url":37,"@type":38,"width":39,"height":40},"https://docshare.wps.com/thumbnails/a-neural-representation-of-sketch-drawings/664823.png","ImageObject",300,407,{"name":42,"@type":43},"acoobca","Person",{"url":19,"name":45,"@type":46},"DocShare","Organization","application/pdf","2026-10-29",true,{"@type":51,"interactionType":52,"userInteractionCount":4},"InteractionCounter",{"@type":53},"ViewAction",{"@type":55,"mainEntity":56},"FAQPage",[57,63,67],{"name":58,"@type":59,"acceptedAnswer":60},"What is sketch-rnn and what does it generate?","Question",{"text":61,"@type":62},"Sketch-RNN is an RNN trained to construct stroke-based drawings of common objects. It outputs sketches in a vector format represented as sequences of pen strokes.","Answer",{"name":64,"@type":59,"acceptedAnswer":65},"How is the model trained?",{"text":66,"@type":62},"The model is trained on thousands of crude human-drawn images covering hundreds of object classes. The sketches are encoded as discrete pen-stroke lists rather than pixel grids.",{"name":68,"@type":59,"acceptedAnswer":69},"What generation modes does the paper cover?",{"text":70,"@type":62},"It outlines both unconditional and conditional sketch generation. For conditional generation, it explores the learned latent space and uses latent priors to produce more coherent images.","https://schema.org",{"og:url":32,"og:type":73,"og:title":10,"og:site_name":45,"og:description":12},"article",{"robots":75,"canonical":32},"index,follow",{"doc_id":77,"site_id":7},664823,1793274263,{"code":4,"msg":80,"data":81},"success",[82,86,90,94,99,104,109,113,118,121,125],{"id":22,"doc_module":4,"doc_module_name":25,"category_name":83,"show_sort_weight":84,"slug":85},"Story & Novel",90,"story-novel",{"id":26,"doc_module":4,"doc_module_name":25,"category_name":87,"show_sort_weight":88,"slug":89},"Literature",80,"literature",{"id":33,"doc_module":4,"doc_module_name":25,"category_name":91,"show_sort_weight":92,"slug":93},"Exam",70,"exam",{"id":95,"doc_module":4,"doc_module_name":25,"category_name":96,"show_sort_weight":97,"slug":98},5,"Comic",60,"comic",{"id":100,"doc_module":4,"doc_module_name":25,"category_name":101,"show_sort_weight":102,"slug":103},6,"Technology",50,"technology",{"id":105,"doc_module":4,"doc_module_name":25,"category_name":106,"show_sort_weight":107,"slug":108},7,"Healthcare",40,"healthcare",{"id":110,"doc_module":4,"doc_module_name":25,"category_name":29,"show_sort_weight":111,"slug":112},8,30,"research-report",{"id":114,"doc_module":4,"doc_module_name":25,"category_name":115,"show_sort_weight":116,"slug":117},9,"Religion & Spirituality",20,"religion-spirituality",{"id":116,"doc_module":4,"doc_module_name":25,"category_name":119,"show_sort_weight":116,"slug":120},"World Cup","world-cup",{"id":122,"doc_module":4,"doc_module_name":25,"category_name":123,"show_sort_weight":122,"slug":124},10,"Lifestyle","lifestyle",{"id":126,"doc_module":4,"doc_module_name":25,"category_name":127,"show_sort_weight":95,"slug":128},19,"General","general",{"code":4,"msg":80,"data":130},{"doc_id":77,"user_id":131,"nickname":42,"user_avatar":132,"doc_module":4,"category_id":110,"category_name":29,"doc_title":10,"doc_description":12,"doc_content":133,"file_id":134,"file_url":135,"file_type":136,"file_size":137,"view_count":4,"is_deleted":4,"is_public":22,"is_downloadable":22,"audit_status":22,"page_count":116,"language":138,"language_code":8,"site_id":7,"html_lang":8,"table_of_contents":139,"faqs":140,"seo_title":141,"seo_description":12,"update_tm":78,"read_time":102},3985747870733,"https://ap-avatar.wpscdn.com/davatar_a8503ba1806abce46bf441b54a3ca4cd","A Neural Representation of Sketch Drawings  \nDavid Ha  \nGoogle Brain [hadavid@google.com](hadavid@google.com)  \nDouglas Eck  \nGoogle Brain [deck@google.com](deck@google.com)  \narXiv : 1704 .03477v2 [ cs .NE] 16 Apr 2017  \nAbstract  \nWe present sketch-rnn, a recurrent neural network (RNN) able to construct stroke-based drawings of common objects. The model is trained on thousands of crude human-drawn images representing hundreds of classes. We outline a framework for conditional and unconditional sketch generation, and describe new robust training methods for generating coherent sketch drawings in a vector format.  \n1 Introduction  \nRecently, there have been major advancements in generative modelling of images using neural networks as a generative tool. Generative Adversarial Networks (GANs) [7], Variational Inference (VI) [19], and Autoregressive (AR) [24] models have become popular tools in this fast growing area. Most of the work, however, has been targeted towards modelling rasterized images represented asa two dimensional grid of pixel values. While these models are currently able to generate realistic, low resolution pixel images, a key challenge for many of these models is to generate images with coherent structure. For example, these models may produce amusing images of cats with three or more eyes, or dogs with multiple heads [7] . In this paper we investigate a much lower-dimensional vector-based representation inspired by how people draw.  \nFigure 1: Latent space of cats and firetrucks generated by sketch-rnn.  \nAs humans, we do not understand the world as a grid of pixels, but rather develop abstract concepts to represent what we see. From a young age, we develop the ability to communicate what we see by drawing on paper with a pencil or crayon. In this way we learn to represent an image based on a short sequence of strokes. For example, an archetypical child-drawn house – i.e. a triangle on top of  \na square with door and window added – can be achieved with a few line strokes. Drawings like this may not resemble reality as captured by a photograph, but they do tell us something about how people represent and reconstruct images of the world around them. Objects such as man, woman, eyes, face, cat, dog, etc. can be clearly communicated with simple drawings. Arguably even emotions can be conveyed via line drawings made from only a few strokes. Finally, abstract visual communication is a key part of how humans communicate with each other, with many writing systems composed of symbols that originated from simple drawings known as pictograms.  \nOur goal is to train machines to learn to draw and generalize abstract concepts in a manner similar to humans. As a first step towards this goal, in this work we train neural networks to learn from adataset of crude hand drawn sketches. The sketches are represented in a vector format, not as pixels, and are encoded as a list of discrete pen strokes. By training a generative recurrent neural network to model discrete pen strokes, we can observe the limits of its ability to learn a generative representation of an object that is both compact (only a few pen strokes) and in some ways abstract in that these stroke sequences capture many different strategies for drawing the same object. For example, see Figure 1 (Left), where the model learns several very different ways to draw a cat.  \nThis paper makes the following contributions: We outline a framework for both unconditional and conditional generation of vector images composed of a sequence of lines. Our recurrent neural network-based generative model is capable of producing sketches of common objects in a vector format. We develop a training procedure unique to vector images to make the training more robust. In the conditional generation model, we explore the latent space developed by the model to represent a vector image. We demonstrate that enforcing a prior on the latent space helps the model generate more coherent images. We also d","cbCaicFyDmCSOWIM","https://ap.wps.com/l/cbCaicFyDmCSOWIM","pdf",5598696,"English","# Abstract\n# 1 Introduction\n# 2 Related Work","[{\"question\":\"What is sketch-rnn and what does it generate?\",\"answer\":\"Sketch-RNN is an RNN trained to construct stroke-based drawings of common objects. It outputs sketches in a vector format represented as sequences of pen strokes.\"},{\"question\":\"How is the model trained?\",\"answer\":\"The model is trained on thousands of crude human-drawn images covering hundreds of object classes. The sketches are encoded as discrete pen-stroke lists rather than pixel grids.\"},{\"question\":\"What generation modes does the paper cover?\",\"answer\":\"It outlines both unconditional and conditional sketch generation. For conditional generation, it explores the learned latent space and uses latent priors to produce more coherent images.\"}]","A Neural Representation of Sketch Drawings | PDF"]