Research map / Detection and segmentation / CNN features
AdaEraser: Training-Free Object Removal via Adaptive Attention Suppression
Dingming Liu
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 |
|---|---|---|---|
| replace full self-attention suppression with token-wise adaptive suppression | AttentiveEraser · Mulan | FID: 54.040 -> 51.108 | Improved |
| replace full suppression with adaptive suppression | AttentiveEraser · OABench | FID: 40.373 -> 38.472 | Improved |
| use training-free AdaEraser instead of training-based RORem | SDXL · Mulan | Average Human Ranking: 6.22 ± 0.19 -> 7.08 ± 0.34 | Improved |
| use token-wise presence score instead of timestep-based linear decay (ablation) | AdaEraser · OABench | FID: 38.831 -> 38.472 | Improved |
| use token-wise presence score instead of region-based score (ablation) | AdaEraser · OABench | FID: 38.945 -> 38.472 | Improved |
| use reference at matched timestep x_t^src instead of fixed x_1^src (ablation) | AdaEraser · OABench | CFD: 0.2658 -> 0.2450 | Improved |
| use matched-timestep reference instead of fixed x_T^src (ablation) | AdaEraser · OABench | ReMOVE: 0.8241 -> 0.8316 | Improved |
| use matched-timestep reference instead of fixed x_T/2^src (ablation) | AdaEraser · OABench | FID: 38.713 -> 38.472 | Improved |
| add presence score estimation (extra parallel branch) over AttentiveEraser | AdaEraser · Efficiency | inference time (s): 13.98 -> 15.41 | Hurt |
| add presence score estimation over AttentiveEraser | AdaEraser · Efficiency | memory (MiB): 7966 -> 9014 | Hurt |
Closest papers by meaning
- Attentive Eraser: Unleashing Diffusion Model’s Object Removal Potential via Self-Attention Redirection Guidance (AAAI 2025 · similarity 0.89)
- MiniMax-Remover: Taming Bad Noise Helps Video Object Removal (NeurIPS 2025 · similarity 0.87)
- RORem: Training a Robust Object Remover with Human-in-the-Loop (CVPR 2025 · similarity 0.86)
- SmartEraser: Remove Anything from Images using Masked-Region Guidance (CVPR 2025 · similarity 0.86)
- Skip-Attention: Improving Vision Transformers by Paying Less Attention (ICLR 2024 · similarity 0.84)
- Detail-Preserving Latent Diffusion for Stable Shadow Removal (CVPR 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.
