Research map / Detection and segmentation / Open-set and few-shot detection
Towards Real Zero-Shot Camouflaged Object Segmentation Without Camouflaged Annotations
Cheng Lei, Jie Fan, Xinran Li, Tian-Zhu Xiang, Ao Li, Ce Zhu, Le Zhang
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use learned codebook instead of M-LLM captions at inference (ablation) | CAMF · COD10K (zero-shot) | weighted F-measure: .708 -> .717 | Improved |
| use learned codebook instead of M-LLM captions at inference (ablation) | CAMF · COD10K (supervised) | weighted F-measure: .826 -> .845 | Improved |
| add multi-scale with MLP projector and mixer versus linear without multi-scale (ablation) | CAMF · COD10K (zero-shot) | weighted F-measure: .669 -> .717 | Improved |
| add multi-scale to linear projector (ablation) | CAMF · CAMO (zero-shot) | weighted F-measure: .664 -> .673 | Mixed |
| use Adapter instead of LoRA as PEFT (ablation) | CAMF · COD10K (zero-shot) | weighted F-measure: .692 -> .717 | Improved |
| use Adapter instead of LoRA as PEFT (ablation) | CAMF · COD10K (supervised) | weighted F-measure: .824 -> .845 | Improved |
| increase MFA channels from 64 to 256 (ablation) | CAMF · COD10K (zero-shot) | weighted F-measure: .691 -> .717 | Improved |
| increase MFA channels from 256 to 512 (ablation) | CAMF · COD10K (zero-shot) | weighted F-measure: .717 -> .498 | Hurt |
| increase MFA channels from 256 to 512 (ablation) | CAMF · COD10K (supervised) | weighted F-measure: .845 -> .841 | No clear effect |
| use prompt 'Describe the image.' instead of 'The object.' for SOS training (ablation) | CAMF · CAMO (zero-shot) | weighted F-measure: .697 -> .729 | Improved |
| use 'Describe the camouflaged animal.' instead of 'The camouflaged object.' in supervised training (ablation) | CAMF · CAMO (supervised) | weighted F-measure: .866 -> .887 | Improved |
| replace M-LLM-based pipelines with codebook inference (CAMF vs GenSAM) | CAMF · inference time | s/img: 3.679 -> 0.055 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- ZoomNeXt — ZoomNeXt: A Unified Collaborative Pyramid Network for Camouflaged Object Detection (TPAMI 2024) · better on 3 of 3 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Discover, Segment, and Select: A Progressive Mechanism for Zero-shot Camouflaged Object Segmentation (CVPR 2026 · similarity 0.91)
- Open-Vocabulary Camouflaged Object Segmentation (ECCV 2024 · similarity 0.89)
- Training-Free Open-Vocabulary Camouflaged Object Segmentation via Fine-Grained Object Binding and Adaptive Hybrid Prompt (CVPR 2026 · similarity 0.89)
- Iterative Refinement of Semantic and Spatial Representations for Open-Vocabulary Camouflaged Object Segmentation (ECCV 2026 · similarity 0.89)
- Shift the Lens: Environment-Aware Unsupervised Camouflaged Object Detection (CVPR 2025 · similarity 0.87)
- Camouflage Anything: Learning to Hide using Controlled Out-painting and Representation Engineering (CVPR 2025 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
