Research map / Detection and segmentation / Weakly supervised detection and segmentation
Learning Camouflaged Object Detection from Noisy Pseudo Label
Jin Zhang, Ruiheng Zhang, Yanjiao Shi, Zhe Cao, Nian Liu, Fahad Shahbaz Khan
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 box annotations as prompts for pseudo-label generation instead of scribbles (ablation) | ANet · pseudo-label quality on remaining 80% of train | F-measure: 0.802 -> 0.860 | Improved |
| Add box prompts versus no prompts (ablation) | ANet · pseudo-label quality on remaining 80% of train | F-measure: 0.751 -> 0.860 | Improved |
| Add point prompts versus no prompts (ablation) | ANet · pseudo-label quality on remaining 80% of train | F-measure: 0.751 -> 0.755 | No clear effect |
| Add box-labeled pseudo-labeled data to 20% full labels instead of unlabeled data (ablation) | PNet · COD10K | F-measure: 0.771 -> 0.792 | Improved |
| Use WSSCOD (20% full + 80% box) versus 20% full only (ablation) | PNet · COD10K | F-measure: 0.764 -> 0.792 | Improved |
| Switch q from 2 to 1 during training (noise correction) instead of fixed q=2 (ablation) | L_NC loss · COD10K | F-measure: 0.778 -> 0.792 | Improved |
| Switch q from 1 to 2 early then 1 versus MAE-form only (ablation) | L_NC loss · COD10K | F-measure: 0.780 -> 0.792 | Improved |
| Use L_NC instead of GCE noise-robust loss (ablation) | PNet · COD10K | F-measure: 0.759 -> 0.792 | Improved |
| Replace the loss with L_NC in SCOD | SCOD · COD10K | F-measure: 0.628 -> 0.665 | Improved |
| Replace the loss with L_NC in SINetv2 | SINetv2 · COD10K | F-measure: 0.719 -> 0.734 | Improved |
| Replace the loss with L_NC in SINetv2 | SINetv2 · COD10K | S-measure: 0.814 -> 0.814 | No clear effect |
| Add 6665 box-only images to training | PNet_F20 · COD10K | S-measure: 0.860 -> 0.901 | Improved |
Closest papers by meaning
- SAM-COD: SAM-guided Unified Framework for Weakly-Supervised Camouflaged Object Detection (ECCV 2024 · similarity 0.92)
- Just a Hint: Point-Supervised Camouflaged Object Detection (ECCV 2024 · similarity 0.92)
- UCOD-DPL: Unsupervised Camouflaged Object Detection via Dynamic Pseudo-label Learning (CVPR 2025 · similarity 0.89)
- Shift the Lens: Environment-Aware Unsupervised Camouflaged Object Detection (CVPR 2025 · similarity 0.89)
- CamoDiffusion: Camouflaged Object Detection via Conditional Diffusion Models (AAAI 2024 · 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.
