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

TL;DR

This paper introduces a hierarchical residual network with a novel compact triplet-center loss to improve sketch recognition accuracy, effectively capturing multi-scale features and addressing intra- and inter-class space issues.

Contribution

It proposes a multi-scale residual block, a hierarchical residual structure, and a specialized triplet-center loss tailored for sketch recognition, advancing feature discrimination.

Findings

Outperforms most baseline methods on Tu-Berlin benchmark

Effective in capturing multi-scale sketch features

Shows superiority among non-sequential models

Abstract

With the widespread use of touch-screen devices, it is more and more convenient for people to draw sketches on screen. This results in the demand for automatically understanding the sketches. Thus, the sketch recognition task becomes more significant than before. To accomplish this task, it is necessary to solve the critical issue of improving the distinction of the sketch features. To this end, we have made efforts in three aspects. First, a novel multi-scale residual block is designed. Compared with the conventional basic residual block, it can better perceive multi-scale information and reduce the number of parameters during training. Second, a hierarchical residual structure is built by stacking multi-scale residual blocks in a specific way. In contrast with the single-level residual structure, the learned features from this structure are more sufficient. Last but not least, the compact triplet-center loss is proposed specifically for the sketch recognition task. It can solve the problem that the triplet-center loss does not fully consider too large intra-class space and too small inter-class space in sketch field. By studying the above modules, a hierarchical residual network as a whole is proposed for sketch recognition and evaluated on Tu-Berlin benchmark thoroughly. The experimental results show that the proposed network outperforms most of baseline methods and it is excellent among non-sequential models at present.