Research map / Detection and segmentation / Weakly supervised detection and segmentation
Image Manipulation Detection With Implicit Neural Representation and Limited Supervision
Zhenfei Zhang, Mingyang Li, Xin Li, Ming-Ching Chang, Jun-Wei Hsieh
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 INR reconstruction error map (NRR) as input (ablation) | FCN baseline, weakly supervised · CASIAv1 | P-F1: 0.122 -> 0.159 | Improved |
| add selective pixel-level contrastive learning (ablation) | Baseline+NRR · CASIAv1 | P-F1: 0.159 -> 0.166 | Improved |
| add adaptive global average pooling (ablation) | Baseline+NRR+SCL · CASIAv1 | P-F1: 0.166 -> 0.199 | Improved |
| add NRR, SCL and AGAP to FCN baseline (ablation) | FCN · NIST16 | P-F1: 0.052 -> 0.131 | Improved |
| add pseudo-label selection to self-distillation (ablation) | unsupervised IMD · NIST16 | P-F1: 0.082 -> 0.129 | Improved |
| use adaptive global-average pooling instead of global-max pooling (ablation) | weakly supervised IMD · CASIAv1 | C-F1: 0.062 -> 0.310 | Improved |
| use adaptive pooling instead of global-average pooling (ablation) | weakly supervised IMD · CASIAv1 | C-F1: 0.195 -> 0.310 | Improved |
| use weakly supervised method instead of WSCL | weakly supervised IMD · Columbia | I-F1: 0.720 -> 0.695 | Hurt |
| train with image-level labels on 12.5K images versus fully supervised methods trained on large synthetic data | weakly supervised IMD · MagicBrush novel manipulation | C-F1: 0.382 vs 0.418 (Trufor), 0.324 (PSCC-Net) | Mixed |
| INR-reconstructed test images evaluated by fully supervised methods | fully supervised IMD methods · CASIAv1, Coverage, Columbia | pixel-level F1 | Hurt |
Closest papers by meaning
- Training-Free Image Manipulation Localization Using Diffusion Models (AAAI 2025 · similarity 0.87)
- A New Benchmark and Model for Challenging Image Manipulation Detection (AAAI 2024 · similarity 0.87)
- M²RL-Net: Multi-View and Multi-Level Relation Learning Network for Weakly-Supervised Image Forgery Detection (AAAI 2025 · similarity 0.86)
- IMDPrompter: Adapting SAM to Image Manipulation Detection by Cross-View Automated Prompt Learning (ICLR 2025 · similarity 0.85)
- UnionFormer: Unified-Learning Transformer with Multi-View Representation for Image Manipulation Detection and Localization (CVPR 2024 · similarity 0.85)
- IDseq: Decoupled and Sequentially Detecting and Grounding Multi-Modal Media Manipulation (AAAI 2025 · similarity 0.85)
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.
