Lightweight Diffusion Model for Camouflaged Object Detection

Camouflaged Object Detection (COD) is an important task in the field of computer vision, which refers to the process of identifying and segmenting objects that seamlessly blend with their surroundings. Applying diffusion models to camouflaged object detection can yield more precise detection results. However, due to the large network architecture and multiple iteration steps of diffusion models, the training speed of these models tends to be slow. This paper proposes a lightweight method for camouflaged object detection based on diffusion models. Firstly, a lightweight U-shaped network is designed. Subsequently, a lightweight Fusion Upsample Enhancement (FUE) module is introduced to enhance, refine, and upsample the extracted features. Experiments show that the proposed LDiffCOD network, when compared with 10 other benchmark algorithms on two COD datasets, achieves the best performance or equal to the optimal results across the board.