ZoomNeXt: A Unified Collaborative Pyramid Network for Camouflaged Object Detection

31 Oct 2023  ·  Youwei Pang, Xiaoqi Zhao, Tian-Zhu Xiang, Lihe Zhang, Huchuan Lu ·

Recent camouflaged object detection (COD) attempts to segment objects visually blended into their surroundings, which is extremely complex and difficult in real-world scenarios. Apart from the high intrinsic similarity between camouflaged objects and their background, objects are usually diverse in scale, fuzzy in appearance, and even severely occluded. To this end, we propose an effective unified collaborative pyramid network which mimics human behavior when observing vague images and videos, \textit{i.e.}, zooming in and out. Specifically, our approach employs the zooming strategy to learn discriminative mixed-scale semantics by the multi-head scale integration and rich granularity perception units, which are designed to fully explore imperceptible clues between candidate objects and background surroundings. The former's intrinsic multi-head aggregation provides more diverse visual patterns. The latter's routing mechanism can effectively propagate inter-frame difference in spatiotemporal scenarios and adaptively ignore static representations. They provides a solid foundation for realizing a unified architecture for static and dynamic COD. Moreover, considering the uncertainty and ambiguity derived from indistinguishable textures, we construct a simple yet effective regularization, uncertainty awareness loss, to encourage predictions with higher confidence in candidate regions. Our highly task-friendly framework consistently outperforms existing state-of-the-art methods in image and video COD benchmarks. The code will be available at \url{https://github.com/lartpang/ZoomNeXt}.

PDF Abstract

Results from the Paper


Task Dataset Model Metric Name Metric Value Global Rank Uses Extra
Training Data
Result Benchmark
Camouflaged Object Segmentation CAMO ZoomNeXt-ResNet-50 MAE 0.065 # 7
Weighted F-Measure 0.774 # 7
S-Measure 0.833 # 6
Camouflaged Object Segmentation CAMO ZoomNeXt-PVTv2-B4 MAE 0.04 # 2
Weighted F-Measure 0.859 # 2
S-Measure 0.888 # 3
Camouflaged Object Segmentation CAMO ZoomNeXt-PVTv2-B5 MAE 0.041 # 3
Weighted F-Measure 0.857 # 3
S-Measure 0.889 # 2
Camouflaged Object Segmentation Camouflaged Animal Dataset ZoomNeXt-PVTv2-B5 S-measure 0.757 # 1
weighted F-measure 0.593 # 1
MAE 0.020 # 1
mDice 0.599 # 1
mIoU 0.510 # 1
Camouflaged Object Segmentation CHAMELEON ZoomNeXt-PVTv2-B4 S-measure 0.925 # 2
weighted F-measure 0.897 # 2
MAE 0.016 # 2
Camouflaged Object Segmentation CHAMELEON ZoomNeXt-PVTv2-B5 S-measure 0.924 # 3
weighted F-measure 0.885 # 3
MAE 0.018 # 3
Camouflaged Object Segmentation CHAMELEON ZoomNeXt-ResNet-50 S-measure 0.908 # 4
weighted F-measure 0.858 # 4
MAE 0.021 # 4
Camouflaged Object Segmentation COD ZoomNeXt-ResNet-50 MAE 0.026 # 3
Weighted F-Measure 0.768 # 4
S-Measure 0.861 # 4
Camouflaged Object Segmentation COD ZoomNeXt-PVTv2-B4 MAE 0.017 # 1
Weighted F-Measure 0.838 # 2
S-Measure 0.898 # 2
Camouflaged Object Segmentation COD ZoomNeXt-PVTv2-B5 MAE 0.018 # 2
Weighted F-Measure 0.827 # 3
S-Measure 0.898 # 2
Camouflaged Object Segmentation MoCA-Mask ZoomNeXt-PVTv2-B5 S-measure 0.734 # 1
weighted F-measure 0.476 # 1
MAE 0.010 # 1
mDice 0.497 # 1
mIoU 0.422 # 1
Camouflaged Object Segmentation NC4K ZoomNeXt-ResNet-50 S-measure 0.874 # 4
weighted F-measure 0.816 # 4
MAE 0.037 # 4
Camouflaged Object Segmentation NC4K ZoomNeXt-PVTv2-B5 S-measure 0.903 # 2
weighted F-measure 0.863 # 3
MAE 0.028 # 2
Camouflaged Object Segmentation NC4K ZoomNeXt-PVTv2-B4 S-measure 0.900 # 3
weighted F-measure 0.865 # 2
MAE 0.028 # 2

Methods


No methods listed for this paper. Add relevant methods here