Research map / Detection and segmentation / Open-set and few-shot detection
Sticking Information in Plain Sight: Encoding and Detecting Hidden Stickers in the Real World
Christina Shatford, Szymon Rusinkiewicz
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 |
|---|---|---|---|
| add total variation loss (ablation) | Ours · ADE20K test (augmented) | decode accuracy (%): LPIPS 0.0632 -> 0.0428; decode 89.16% -> 84.20% | Mixed |
| add simulated placement jitter in training (ablation) | Ours · ADE20K test (augmented) | decode accuracy (%): 86.85% -> 89.16% | Improved |
| use LPIPS instead of MSE for cover loss (ablation) | Ours · ADE20K test (augmented) | LPIPS: 0.1111 -> 0.0632 | Improved |
| add noise augmentation (ablation) | Ours · ADE20K test (augmented) | detection rate (%): 46.45% -> 97.24% | Improved |
| add noise augmentation (ablation) | Ours · ADE20K test (augmented) | LPIPS: 0.0125 -> 0.0632 | Hurt |
| allow color patterns instead of greyscale (ablation) | Ours · ADE20K test (augmented) | LPIPS: 0.2132 -> 0.0632 | Improved |
| add L_predLocDecode loss (ablation) | Ours · ADE20K test (augmented) | decode accuracy (%): approximately 6.4 percentage points | Improved |
| emphasize decoding accuracy over visibility (ablation) | Ours · ADE20K test (augmented) | decode accuracy (%): 89.16% -> 94.80%; LPIPS 0.0632 -> 0.1099 | Mixed |
| emphasize low visibility over accuracy (ablation) | Ours · ADE20K test (augmented) | LPIPS: LPIPS 0.0632 -> 0.0290; decode 89.16% -> 81.93% | Mixed |
| make stickers universal rather than location-specific (ablation) | Ours · ADE20K test (augmented) | decode accuracy (%): 89.16% -> 87.89%; LPIPS 0.0632 -> 0.0667 | Mixed |
| use sticker with placement-jitter-aware training instead of StegaStamp | StegaStamp · 400x400 container | PSNR (dB): 28.60 -> 32.27 | Improved |
| use our sticker pipeline instead of StegaStamp under 9 px misalignment | StegaStamp · printed real-world | decode accuracy (%): 99.85% -> 73.00% for StegaStamp at 0 to 9 px | Improved |
Closest papers by meaning
- Learning Invisible Markers for Hidden Codes in Offline-to-Online Photography (CVPR 2022 · similarity 0.83)
- ObjectFormer for Image Manipulation Detection and Localization (CVPR 2022 · similarity 0.82)
- Identity From Here, Pose From There: Self-Supervised Disentanglement and Generation of Objects Using Unlabeled Videos (ICCV 2019 · similarity 0.81)
- FakeShield: Explainable Image Forgery Detection and Localization via Multi-modal Large Language Models (ICLR 2025 · similarity 0.80)
- Physical Marker: Revealing Invisible Hyperlinks Hidden in Printed Trademarks (AAAI 2025 · similarity 0.80)
- Seeing the Unseen: Visual Common Sense for Semantic Placement (CVPR 2024 · similarity 0.79)
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.
