Research map / Image generation / Autoregressive and latent generation
Learning JPEG Compression Artifacts for Image Manipulation Detection and Localization
Myung-Joon Kwon, Seung-Hun Nam, In-Jae Yu, Heung-Kyu Lee, Changick Kim
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 double JPEG pretraining of the DCT stream (ablation) | CAT-Net · GRIP (copy-move) | p-F1: 24.60 -> 76.45 | Improved |
| add double JPEG pretraining of the DCT stream (ablation) | CAT-Net · GRIP (copy-move) | p-AP: 43.43 -> 91.87 | Improved |
| add double JPEG pretraining of the DCT stream (ablation) | CAT-Net · Carvalho (splicing) | p-F1: 51.33 -> 78.79 | Improved |
| add double JPEG pretraining of the DCT stream (ablation) | CAT-Net · Columbia (no initial compression) | p-F1: 95.15 -> 93.97 | Hurt |
| add double JPEG pretraining of the DCT stream (ablation) | CAT-Net · Columbia (no initial compression) | p-AP: 98.21 -> 95.87 | Hurt |
| add double JPEG pretraining of the DCT stream (ablation) | CAT-Net · CoMoFoD (copy-move) | p-F1: 15.53 -> 14.01 | Hurt |
| add the DCT stream to the RGB stream (ablation) | CAT-Net · GRIP (copy-move) | p-F1: 9.46 -> 76.45 | Improved |
| add the RGB stream to the DCT stream (ablation) | CAT-Net · Columbia (splicing) | p-F1: 71.88 -> 93.97 | Improved |
| add the DCT stream to the RGB stream (ablation) | CAT-Net · CoMoFoD (copy-move) | p-F1: 17.48 -> 14.01 | Hurt |
| add DCT volume representation instead of raw DCT input (ablation) | ResNet152 · Double JPEG detection | accuracy: 54.08 -> 90.19 | Improved |
| add the quantization table (QT) input to the DCT stream (ablation) | DCT Stream · Double JPEG detection | accuracy: 91.71 -> 93.93 | Improved |
| use the DCT stream architecture instead of ResNet152 on DCT volume | DCT volume input · Double JPEG detection | accuracy: 90.19 -> 91.71 | Improved |
Closest papers by meaning
- JPEG Artifacts Reduction via Deep Convolutional Sparse Coding (ICCV 2019 · similarity 0.88)
- Self-Supervised Domain Adaptation for Forgery Localization of JPEG Compressed Images (ICCV 2021 · similarity 0.86)
- Can We Model the Artifacts Explicitly? Disentangle Artifacts via Pairwise Edit Relations for Image Manipulation Localization (NeurIPS 2026 · similarity 0.86)
- ManTra-Net: Manipulation Tracing Network for Detection and Localization of Image Forgeries With Anomalous Features (CVPR 2019 · similarity 0.86)
- Towards JPEG-Resistant Image Forgery Detection and Localization Via Self-Supervised Domain Adaptation (TPAMI 2022 · similarity 0.86)
- A New Benchmark and Model for Challenging Image Manipulation Detection (AAAI 2024 · similarity 0.84)
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.
