Dual Skipping Networks

TL;DR

This paper introduces a dual skipping network inspired by neuroscience, enabling simultaneous coarse and fine object classification through a layer-skipping mechanism that predicts which layers to skip, demonstrating promising results on benchmark datasets.

Contribution

The paper proposes a novel dual skipping network with a learned gating mechanism for flexible layer skipping, enhancing coarse-to-fine object categorization.

Findings

Achieves promising results on benchmark datasets

Demonstrates effective coarse-to-fine classification

Introduces a flexible layer-skipping mechanism

Abstract

Inspired by the recent neuroscience studies on the left-right asymmetry of the human brain in processing low and high spatial frequency information, this paper introduces a dual skipping network which carries out coarse-to-fine object categorization. Such a network has two branches to simultaneously deal with both coarse and fine-grained classification tasks. Specifically, we propose a layer-skipping mechanism that learns a gating network to predict which layers to skip in the testing stage. This layer-skipping mechanism endows the network with good flexibility and capability in practice. Evaluations are conducted on several widely used coarse-to-fine object categorization benchmarks, and promising results are achieved by our proposed network model.