Research map / Image generation / Image restoration
Learning Invisible Markers for Hidden Codes in Offline-to-Online Photography
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Learn invisible markers and localize the encoded sub-image automatically | Ours · Geometric distortion [-20,+20] to [-40,+40] | BER: 2.04%, 4.01%, 4.74%, 7.53%, 13.77% vs StegaStamp 14.45%, 24.10%, 27.85%, 32.15%, 36.27% and RIHOOP 48.20%, 51.30%, 48.90 | Helped |
| Hide information in a sub-image and train an end-to-end localization module | Ours · Test images | PSNR: 32.95 vs 29.82 and 35.86 | Mixed |
| Use the localization network instead of manual correction (ablation) | Ours · Photograph preprocessing | time: 0.014s vs 7s | Helped |
| Use the localization network instead of manual correction (ablation) | Ours · Distortion 1 and Distortion 2 | BER: 0.16% -> 2.10%; 1.22% -> 7.57% | Hurt |
| Add the edge gain loss (ablation) | Ours · Test images | IoU: 0.8866 -> 0.9318 | Helped |
| Add the edge gain loss (ablation) | Ours · Test images | BER: 11.29% -> 8.08% | Helped |
| Add the edge gain loss (ablation) | Ours · Test images | PSNR: 40.64 -> 41.87 | Helped |
| Hide information in high-frequency regions instead of low-frequency regions (ablation) | Ours · Different sub-image selections | BER: BER increases by 9.35% | Hurt |
| Train with a multi-stage strategy rather than optimizing all losses at once (ablation) | Ours · Training | convergence | Helped |
Closest papers by meaning
- Physical Marker: Revealing Invisible Hyperlinks Hidden in Printed Trademarks (AAAI 2025 · similarity 0.84)
- Robust Message Embedding via Attention Flow-Based Steganography (CVPR 2025 · similarity 0.83)
- WMarkGPT: Watermarked Image Understanding via Multimodal Large Language Models (ICML 2025 · similarity 0.82)
- FreqMark: Invisible Image Watermarking via Frequency Based Optimization in Latent Space (NeurIPS 2024 · similarity 0.82)
- An Undetectable Watermark for Generative Image Models (ICLR 2025 · similarity 0.81)
- Hidden in the Noise: Two-Stage Robust Watermarking for Images (ICLR 2025 · similarity 0.81)
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.