Integrating part-object relationship and contrast for camouflaged object detection

Abstract

Object detectors that solely rely on image contrast are struggling to detect camouflaged objects in images because of the high similarity between camouflaged objects and their surroundings. To address this issue, in this paper, we investigate the role of the part-object relationship for camouflaged object detection.

Specifically, we propose a Part-Object relationship and Contrast Integrated Network (POCINet) covering both search and identification stages, where each stage adopts an appropriate scheme to engage the contrast information and part-object relational knowledge for camouflaged pattern decoding.

Methodology

POCINet Framework: Our network covers both search and identification stages, with each stage utilizing contrast information and part-object relational knowledge for effective camouflaged pattern decoding.

Search-to-Identification Guidance (SIG) Module: We bridge these two stages via the SIG module, in which the search result, as well as decoded semantic knowledge, jointly enhances the features encoding ability of the identification stage.

The framework employs an encoder-decoder architecture with multi-stage processing to progressively refine the detection of camouflaged objects by integrating both low-level contrast cues and high-level part-object relationships.

Experimental Results

Experimental results demonstrate the superiority of our algorithm on three datasets. Notably, our algorithm raises F_β of the best existing method by approximately 17 points on the CPD1K dataset.

The significant performance improvements validate the effectiveness of integrating part-object relationships with contrast information for camouflaged object detection tasks.