A Novel Structure of Convolutional Layers with a Higher Performance-Complexity Ratio for Semantic Segmentation

Yalong Jiang, Zheru Chi

Research output: Chapter in book / Conference proceedingConference article published in proceeding or bookAcademic researchpeer-review

Abstract

In this paper, we study an important factor that determines the capacity of a CNN model and propose a novel structure of convolutional layers with a higher performance-complexity ratio. Firstly, the relationship of the model capacity and the number of parameters versus segmentation performance is explored. Secondly, a mechanism is proposed to optimize the structure of a CNN model for a specific task. The mechanism also provides better convergence than current state-of-the-art methods for factorizing convolutional layers, such as MobileNet. Thirdly, we propose a measure based on the mutual information between hidden activations and inputs/outputs to compute the capacity of a CNN model. This measure is highly correlated with segmentation performance. Experimental results on the segmentation of the PASCAL Person Parts Dataset show that the linear dependency among convolutional kernels is an important factor determining the capacity of a CNN model. It is also demonstrated that our approach can successfully adjust the model capacity to best match to the complexity of a dataset. The optimized CNN model achieves the similar performance to Deeplab-V2 on the segmentation task with 100 × less parameters, resulting in a significantly improved performance-complexity ratio.

Original languageEnglish
Title of host publication2018 15th International Conference on Control, Automation, Robotics and Vision, ICARCV 2018
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages186-191
Number of pages6
ISBN (Electronic)9781538695821
DOIs
Publication statusPublished - 18 Nov 2018
Event15th International Conference on Control, Automation, Robotics and Vision, ICARCV 2018 - Singapore, Singapore
Duration: 18 Nov 201821 Nov 2018

Publication series

Name2018 15th International Conference on Control, Automation, Robotics and Vision, ICARCV 2018

Conference

Conference15th International Conference on Control, Automation, Robotics and Vision, ICARCV 2018
Country/TerritorySingapore
CitySingapore
Period18/11/1821/11/18

ASJC Scopus subject areas

  • Artificial Intelligence
  • Computer Vision and Pattern Recognition
  • Control and Optimization

Fingerprint

Dive into the research topics of 'A Novel Structure of Convolutional Layers with a Higher Performance-Complexity Ratio for Semantic Segmentation'. Together they form a unique fingerprint.

Cite this