TPRNet: camouflaged object detection via transformer-induced progressive refinement network

• Published in: The Visual computer Vol. 39; no. 10; pp. 4593 - 4607
• Main Authors: Zhang, Qiao, Ge, Yanliang, Zhang, Cong, Bi, Hongbo
• Format: Journal Article
• Language: English
• Published: Berlin/Heidelberg Springer Berlin Heidelberg 01.10.2023; Springer Nature B.V
• Subjects: Artificial Intelligence; Computer Graphics; Computer Science; Deep learning; Image Processing and Computer Vision; Methods; Mining; Modules; Neural networks; Object recognition; Original Article; Semantics; Transformers; Deep learning; Transformer; Progressive refinement; Camouflaged object detection
• ISSN: 0178-2789; 1432-2315
• DOI: 10.1007/s00371-022-02611-1

Camouflaged object detection (COD) is a challenging task which aims to detect objects similar to the surrounding environment. In this paper, we propose a transformer-induced progressive refinement network ( TPRNet ) to solve challenging COD tasks. Specifically, our network includes a Transformer-induced Progressive Refinement Module (TPRM) and a Semantic-Spatial Interaction Enhancement Module (SIEM). In TPRM, high-level features with rich semantic information are integrated through transformers as prior guidance, and then, it is sent to the refinement concurrency unit (RCU), and the accurately positioned feature area is obtained through a progressive refinement strategy. In SIEM, we perform feature interaction to localized-accurate semantic features and low-level features to obtain rich fine-grained clues and increase the symbolic power of boundary features. Extensive experiments on four widely used benchmark datasets (i.e., CAMO, CHAMELEON, COD10K, and NC4K) demonstrate that our TPRNet is an effective COD model and outperforms state-of-the-art models. The code is available https://github.com/zhangqiao970914/TPRNet .