Research map / Video / Video segmentation and tracking
Beyond Pixels: Mining Compressed Domain Artifacts for Efficient AI-Generated Video Detection
Anran Zhu, Zhengli Shi, Chende Zheng, Chenhao Lin, Zhengyu Zhao, Le Yang, Chong Zhang, Shuai Liu, Chao Shen
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 |
|---|---|---|---|
| fuse I-frame, motion vector and residual branches instead of I-frame only (ablation) | STREAM · GVF | mAP: 0.885 -> 0.950 | Improved |
| use compressed-domain bitstream inputs instead of pixel-domain decoding | AIGV detector · inference time per 1,000 videos | seconds: 40 -> 16 (vs D3) | Improved |
| use compressed-domain STREAM instead of best pixel-domain baseline D3 | AIGV detection · 5-dataset average | mAP: 0.943 -> 0.965 | Improved |
| use ResNet-152 instead of ResNet-18 backbone (ablation) | STREAM · 10 generators | mAP: 0.9149 -> 0.9204 | No clear effect |
| add strong Gaussian blur to test videos (residual only) (ablation) | STREAM residual branch · 10 generators | mAP: 0.941 -> 0.842 | Hurt |
| add strong Gaussian blur to test videos (MV only) (ablation) | STREAM MV branch · 10 generators | mAP: 0.937 -> 0.924 | Hurt |
| fuse all three modalities under strong Gaussian blur (ablation) | STREAM · 10 generators | mAP: 0.975 -> 0.949 | Hurt |
| increase rounds of loop-filter re-encoding from 0 to 6 | STREAM, AIGVDet, NPR · multi-round re-encoding | mAP: 0.976 -> 0.697 (STREAM) | Hurt |
| H.264 re-encoding at QP=35 of test videos | STREAM · VideoPhy/GenVideo/GVF | avg mAP: 0.959 -> 0.902 | Hurt |
| overlay 50% opacity watermark on test videos | STREAM · VideoPhy/GenVideo/GVF | avg mAP: 0.959 -> 0.827 | Hurt |
| change resolution to 720p or re-encode with H.265 | STREAM · VideoPhy/GenVideo/GVF | avg mAP: 0.959 -> 0.958 | No clear effect |
Closest papers by meaning
- Rethinking the Readout: Unlocking Video Backbones for AI-Generated Video Detection (NeurIPS 2026 · similarity 0.90)
- Your One-Stop Solution for AI-Generated Video Detection (CVPR 2026 · similarity 0.87)
- Physics-Driven Spatiotemporal Modeling for AI-Generated Video Detection (NeurIPS 2025 · similarity 0.87)
- HIMMEL: Hierarchical Interleaved Multi-stream Motion Encoding for Long Video Understanding (NeurIPS 2026 · similarity 0.84)
- A Sanity Check for AI-generated Image Detection (ICLR 2025 · similarity 0.84)
- Sparse Video-Gen: Accelerating Video Diffusion Transformers with Spatial-Temporal Sparsity (ICML 2025 · 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.
