The sketchy database

Sangkloy, Patsorn; Burnell, Nathan; Ham, Cusuh; Hays, James

doi:10.1145/2897824.2925954

Cited by 513 publications

(95 citation statements)

References 53 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Most commonly used sketch datasets are Sketchy database [22], TU-Berlin [8], QuickDraw [11] and QMUL Shoe/Chair SBIR dataset [34]. Additionally, the CUHK Face Sketches [28] are drawn by artists using pen and paper.…”

Section: Datasetmentioning

confidence: 99%

See 1 more Smart Citation

SketchMan

Gao

Shen

et al. 2020

Proceedings of the 28th ACM International Conference on Multimedia

View full text Add to dashboard Cite

Human free-hand sketches have been studied in various fields including sketch recognition, synthesis and sketch-based image retrieval. We propose a new challenging task sketch enhancement (SE) defined in an ill-posed space, i.e. enhancing a non-professional sketch (NPS) to a professional sketch (PS), which is a creative generation task different from sketch abstraction, sketch completion and sketch variation. For the first time we release a database of NPS with PS for anime characters. We cast sketch enhancement as an image-to-image translation problem by exploiting the relationship to corresponding intensive or sparse pixel domains for sketch domain. Specifically, we explore three different routines based on conditional generative adversarial network (cGAN), i.e. Sketch-Sketch (SS), Sketch-Colorization-Sketch (SCS) and Sketch-Abstraction-Sketch (SAS). SS is a one-stage model that directly maps NPS to PS, while SCS and SAS are two-stage models where auxiliary inputs, grayscale parsing and shape parsing, are involved. Multiple metrics are used to evaluate the performance of the models in both the sketch domain and other low-level feature domains. With quantitative and qualitative analysis of the experiments, we have established solid baselines, which, we hope, could encourage more research conducted on this task. Our dataset is publicly available via https://github.com/LCXCUC/SketchMan2020.

show abstract

Section: Datasetmentioning

confidence: 99%

“…10 characters with distinct appearances and poses in one scene. The sketches are fairly more complex-structured than that of existing databases such as [8,11,22]. For each sample, there is a free-hand NPS and an approximate professional version corresponding to PS.…”

Section: Introductionmentioning

confidence: 99%

SketchMan

Gao

Shen

et al. 2020

Proceedings of the 28th ACM International Conference on Multimedia

View full text Add to dashboard Cite

show abstract

“…With these premises, we searched for large and public datasets (possibly with an available ground truth) to be used in this evaluation. One possible solution is the use of the SHREC13 -"Testing Sketches" dataset [Li et al 2013], whereas alternatives are Google Quick Draw and Sketchy dataset [Sangkloy et al 2016]. SHREC13 contains very clean, single-stroke "hand-drawn" sketches (created using a touch pad or mouse), such as those reported in Figs.…”

Section: Experiments 41 Line Extractionmentioning

confidence: 99%

A complete hand-drawn sketch vectorization framework

Donati

Cesano

Prati

2019

Multimed Tools Appl

View full text Add to dashboard Cite

Vectorizing hand-drawn sketches is a challenging task, which is of paramount importance for creating CAD vectorized versions for the fashion and creative workflows. This paper proposes a complete framework that automatically transforms noisy and complex hand-drawn sketches with different stroke types in a precise, reliable and highly-simplified vectorized model.The proposed framework includes a novel line extraction algorithm based on a multi-resolution application of Pearson's cross correlation and a new unbiased thinning algorithm that can get rid of scribbles and variable-width strokes to obtain clean 1-pixel lines. Other contributions include variants of pruning, merging and edge linking procedures to post-process the obtained paths. Finally, a modification of the original Schneider's vectorization algorithm is designed to obtain fewer control points in the resulting Bezier splines.All the proposed steps of the framework have been extensively tested and compared with state-of-the-art algorithms, showing (both qualitatively and quantitatively) its outperformance.

show abstract

“…With the recent increasing popularity of neural networks and crowdsourcing, researchers have developed large-scale sketch datasets that correspond to natural image datasets to power neural-network-driven techniques for image-retrieval tasks. The TU-Berlin [8] and Sketchy [22] sketch datasets consist of crowdsourced sketches that are collected from crowdworkers by presenting them the original corresponding natural images. Using these corresponding sketch-image pairs, neural networks are trained to directly encode matching sketches and images to similar low-dimensional outputs.…”

Section: Sketch-based Image Retrieval and Sketch Datasetsmentioning

confidence: 99%

“…During the training phase, we train Swire's deep neural network to generate similar lowdimensional outputs (64-dimensions) for matching pairs of screenshots and sketches, and dissimilar outputs for nonmatching pairs of screenshots and sketches. This training scheme is shown to be useful for sketch-based image retrieval [22]. In the querying phase, we use Swire's trained neural network to encode a user's sketch query and retrieve UIs with the closest output to the user's sketch's output.…”

Section: Deep Neural-network-based User Interface Retrievalmentioning

confidence: 99%