Consensus method improves camouflaged object detection using multiple references
Consensus-Aware Multi-Source Fusion for Reference-Guided Camouflaged Object Detection
Computer Vision and Pattern Recognition
Summary
Camouflaged objects are hard to spot because they blend in with their surroundings. This paper presents a way to better detect these objects by using multiple reference images that hint at what the target looks like. The authors combine information from different deep learning models and references to consensus on what parts of an image belong to the camouflaged object. Their approach improves the accuracy of identifying these hidden objects compared to previous methods.
What this means in practice
- •For wildlife conservation teams: Identify and segment camouflaged animals in natural habitats using multiple visual references to assist monitoring efforts.
- •For security surveillance operators: Detect concealed objects or persons in complex scenes by integrating several reference samples for more reliable identification.
Authors
Junyang Xia, Luocheng Zhang, Wenwen Pan, Chifeng Zhu, Yang Yang, Xinchun Liu, Jiajun Ding
Abstract
Reference-guided camouflaged object detection aims to segment a target whose visual appearance closely resembles its surroundings by exploiting auxiliary reference samples. The task remains difficult because reference samples contain inconsistent target cues, while generic visual representations are not inherently aligned with the target specified by the references. To handle these problems, we present a consensus-aware multi-source fusion framework. Reference-Conditioned Dual-Backbone Fusion (RCDF) couples trainable PVTv2 query features with frozen DINOv3 representations and uses reference-conditioned correlation to select foundation-model evidence before multi-scale fusion. The framework also aggregates multiple references through cross-reference consensus aggregation and injects reference information at semantic depths matched to the query features. Extensive experiments demonstrate the effectiveness of the proposed method. The results further show that reference consensus, target-conditioned foundation features, and hierarchical decoding provide complementary improvements under the evaluation protocol. The source code will be made publicly available upon acceptance.