SFSegNet segments freehand sketches into object parts with a ResNet34-based FCN that ignores background pixels during training and applies affine transform encoders, achieving 91.2% average pixel accuracy on the authors' dataset.
Multi-column Point-CNN for Sketch Segmentation
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
abstract
Traditional sketch segmentation methods mainly rely on handcrafted features and complicate models, and their performance is far from satisfactory due to the abstract representation of sketches. Recent success of Deep Neural Networks (DNNs) in related tasks suggests DNNs could be a practical solution for this problem, yet the suitable datasets for learning and evaluating DNNs are limited. To this end, we introduce SketchSeg, a large dataset consisting of 10,000 pixel-wisely labeled sketches.Besides, due to the lack of colors and textures in sketches, conventional DNNs learned on natural images are not optimal for tackling our problem.Therefore, we further propose the Multi-column Point-CNN (MCPNet), which (1) directly takes sampled points as its input to reduce computational costs, and (2) adopts multiple columns with different filter sizes to better capture the structures of sketches. Extensive experiments validate that the MCPNet is superior to conventional DNNs like FCN. The SketchSeg dataset is publicly available on https://drive.google.com/open?id=1OpCBvkInhxvfAHuVs-spDEppb8iXFC3C.
fields
cs.CV 1years
2019 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
SFSegNet: Parse Freehand Sketches using Deep Fully Convolutional Networks
SFSegNet segments freehand sketches into object parts with a ResNet34-based FCN that ignores background pixels during training and applies affine transform encoders, achieving 91.2% average pixel accuracy on the authors' dataset.