Research map / Video / Video segmentation and tracking
Frequency-Aware Spatiotemporal Transformers for Video Inpainting Detection
Bingyao Yu, Wanhua Li, Xiu Li, Jiwen Lu, Jie Zhou
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 frequency-aware features (DCT band decomposition + CNN branch) (ablation) | FAST · DVI (trained VI+OP, tested VI*/OP*/CP) | mIoU: 0.55/0.51/0.48 -> 0.61/0.65/0.63 | Improved |
| add focal loss to the hybrid loss (ablation) | FAST · DVI (trained VI+OP, tested VI*/OP*/CP) | mIoU: 0.57/0.58/0.59 -> 0.61/0.65/0.63 | Improved |
| add IoU loss to the hybrid loss (ablation) | FAST · DVI (trained VI+OP, tested VI*/OP*/CP) | mIoU: 0.58/0.59/0.59 -> 0.61/0.65/0.63 | Improved |
| use a 3-frame clip instead of a single frame (ablation) | FAST · DVI (trained VI+OP, tested VI*/OP*/CP) | mIoU: 0.57/0.53/0.51 -> 0.61/0.65/0.63 | Improved |
| compare FAST with VIDNet on frame-level inpainting AUC | FAST vs VIDNet · DVI frame-level classification | AUC: 0.778/0.768/0.884 -> 0.795/0.787/0.898 | Improved |
| compare FAST with VIDNet on unseen-method segmentation mIoU | FAST vs VIDNet · DVI unseen methods (CP, VI, OP) | mIoU: CP 0.57 -> 0.63; VI 0.39 -> 0.32; OP 0.25 -> 0.22 | Mixed |
| compare FAST with VIDNet on cross-dataset FVI | FAST vs VIDNet · FVI | mIoU / F1: mIoU 0.257 -> 0.285; F1 0.367 -> 0.359 | Mixed |
| JPEG and Gaussian noise perturbation of test frames vs baselines | FAST · DVI robustness | mIoU | Improved |
Closest papers by meaning
- DLFormer: Discrete Latent Transformer for Video Inpainting (CVPR 2022 · similarity 0.88)
- ProPainter: Improving Propagation and Transformer for Video Inpainting (ICCV 2023 · similarity 0.87)
- WaveFormer: Wavelet Transformer for Noise-Robust Video Inpainting (AAAI 2024 · similarity 0.85)
- Parameterized Stripe Attention for Efficient Video Generation (NeurIPS 2026 · similarity 0.85)
- MotionAura: Generating High-Quality and Motion Consistent Videos using Discrete Diffusion (ICLR 2025 · similarity 0.85)
- CoCoCo: Improving Text-Guided Video Inpainting for Better Consistency, Controllability and Compatibility (AAAI 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.
