We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

cs.CV

Change to browse by:

cs

References & Citations

DBLP - CS Bibliography

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Computer Science > Computer Vision and Pattern Recognition

Title: iSegFormer: Interactive Segmentation via Transformers with Application to 3D Knee MR Images

Abstract: We propose iSegFormer, a memory-efficient transformer that combines a Swin transformer with a lightweight multilayer perceptron (MLP) decoder. With the efficient Swin transformer blocks for hierarchical self-attention and the simple MLP decoder for aggregating both local and global attention, iSegFormer learns powerful representations while achieving high computational efficiencies. Specifically, we apply iSegFormer to interactive 3D medical image segmentation.
Comments: 11 pages, 2 figures, 4 tables
Subjects: Computer Vision and Pattern Recognition (cs.CV)
Cite as: arXiv:2112.11325 [cs.CV]
  (or arXiv:2112.11325v5 [cs.CV] for this version)

Submission history

From: Qin Liu [view email]
[v1] Tue, 21 Dec 2021 16:16:30 GMT (167kb,D)
[v2] Wed, 26 Jan 2022 01:17:22 GMT (167kb,D)
[v3] Mon, 31 Jan 2022 22:27:52 GMT (168kb,D)
[v4] Thu, 3 Mar 2022 04:20:51 GMT (399kb,D)
[v5] Fri, 18 Mar 2022 03:29:48 GMT (1696kb,D)

Link back to: arXiv, form interface, contact.