Abstract:Sketch recognition remains a significant challenge due to the limited training data and the substantial intra-class variance of freehand sketches for the same object. Conventional methods for this task often rely on the availability of the temporal order of sketch strokes, additional cues acquired from different modalities and supervised augmentation of sketch datasets with real images, which also limit the applicability and feasibility of these methods in real scenarios. In this paper, we propose a novel sketch-specific data augmentation (SSDA) method that leverages the quantity and quality of the sketches automatically. From the aspect of quantity, we introduce a Bezier pivot based deformation (BPD) strategy to enrich the training data. Towards quality improvement, we present a mean stroke reconstruction (MSR) approach to generate a set of novel types of sketches with smaller intra-class variances. Both of these solutions are unrestricted from any multi-source data and temporal cues of sketches. Furthermore, we show that some recent deep convolutional neural network models that are trained on generic classes of real images can be better choices than most of the elaborate architectures that are designed explicitly for sketch recognition. As SSDA can be integrated with any convolutional neural networks, it has a distinct advantage over the existing methods. Our extensive experimental evaluations demonstrate that the proposed method achieves the state-of-the-art results (84.27%) on the TU-Berlin dataset, outperforming the human performance by a remarkable 11.17% increase. Finally, more experiments show the practical value of our approach for the task of sketch-based image retrieval.

Light-SRNet: a lightweight dual-attention feature fusion network for hand-drawn sketch recognition

Sketch-R2CNN: an Attentive Network for Vector Sketch Recognition

<i>Sketch-R2CNN</i>: An RNN-Rasterization-CNN Architecture for Vector Sketch Recognition

Sketchformer++: A Hierarchical Transformer Architecture for Vector Sketch Representation

SceneSketcher-v2: Fine-Grained Scene-Level Sketch-Based Image Retrieval Using Adaptive GCNs

A lightweight and stochastic depth residual attention network for remote sensing scene classification

Sketch-a-Net: A Deep Neural Network That Beats Humans

Coupling Deep Textural and Shape Features for Sketch Recognition.

SketchNet: Sketch Classification with Web Images

On Learning Semantic Representations for Million-Scale Free-Hand Sketches

Sketch-a-Net That Beats Humans

Sketchpointnet: A Compact Network for Robust Sketch Recognition

Sketch-R2CNN : An RNN-Rasterization-CNN Architecture for Vector Sketch Recognition

Mixed attention dense network for sketch classification

SFSegNet: Parse Freehand Sketches using Deep Fully Convolutional Networks

Sketch Recognition with Deep Visual-Sequential Fusion Model

A hierarchical residual network with compact triplet-center loss for sketch recognition

Research on Recognition Effect of DSCN Network Structure in Hand-Drawn Sketch

Sketch-Specific Data Augmentation for Freehand Sketch Recognition

Deep Neural Networks for Free-Hand Sketch Recognition.

A Multi-Layer Deep Fusion Convolutional Neural Network for Sketch Based Image Retrieval