Deakin University
Browse

File(s) under embargo

ShuffleTrans: Patch-wise weight shuffle for transparent object segmentation

journal contribution
posted on 2023-10-03, 01:40 authored by B Zhang, Z Wang, Y Ling, Y Guan, S Zhang, W Li, Lei WeiLei Wei, C Zhang
Transparent objects widely exist in the world. The task of transparent object segmentation is challenging as the object lacks its own texture. The cue of shape information therefore gets more critical. Most existing methods, however, rely on the mechanism of simple convolution, which is good at local cues and performs weakly on global cues like shape. To solve this problem, an operation named Patch-wise Weight Shuffle is proposed to bring in the global context cue by being combined with the dynamic convolution. A network ShuffleTrans that recognizes shape better is then designed based on this operation. Besides, fitter for this task, two auxiliary modules are presented in ShuffleTrans: a Boundary and Direction Refinement Module which collects two additional information, and a Channel Attention Enhancement Module that assists the above operation. Experiments on four texture-less object segmentation datasets and two normal datasets verify the effectiveness and generality of the method. Especially, the ShuffleTrans achieved 74.93% mIoU on the Trans10k v2 test set, which is more accurate than existing methods.

History

Journal

Neural Networks

Volume

167

Pagination

199-212

Location

United States

ISSN

0893-6080

eISSN

1879-2782

Language

en

Publication classification

C1.1 Refereed article in a scholarly journal

Publisher

Elsevier BV

Usage metrics

    Research Publications

    Exports

    RefWorks
    BibTeX
    Ref. manager
    Endnote
    DataCite
    NLM
    DC