Research map / Image generation / Editing and style transfer
EraseLoRA: MLLM-Driven Foreground Exclusion and Background Subtype Aggregation for Dataset-Free Object Removal
Sanghyun Jo, Donghwan Lee, Eunji Jung, Seong Je Oh, Kyungsu 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 EraseLoRA (BFE + BRSA) | SD3.5-M · OpenImages V7 (200 samples) | BG Sim.: 0.605 to 0.743 | Improved |
| add EraseLoRA (BFE + BRSA) | SD3.5-M · RORD (343 frames) | BG Sim.: 0.582 to 0.779 | Improved |
| add EraseLoRA (BFE + BRSA) | SD3.5-M · OpenImages V7 (200 samples) | FG Sim.: 0.286->0.151 | Improved |
| add EraseLoRA (BFE + BRSA) | SD3.5-M · RORD (343 frames) | FG Sim.: 0.319->0.138 | Improved |
| add EraseLoRA | SD3.5-M · OpenImages V7 (200 samples) | BG Pres.: 0.934 to 0.924 | No clear effect |
| add BFE to AttentiveEraser (ablation) | AttentiveEraser · OpenImages V7 (200 samples) | BG Sim.: 0.559 to 0.596 | Improved |
| add BFE to AttentiveEraser (ablation) | AttentiveEraser · OpenImages V7 (200 samples) | FG Sim.: 0.276 to 0.252 | Improved |
| add BFE to DesignEdit (ablation) | DesignEdit · OpenImages V7 (200 samples) | BG Sim.: 0.600 to 0.603 | Improved |
| add background puzzle loss to reconstruction loss (ablation) | EraseLoRA · OpenImages V7 (200 samples) | BG Sim.: 0.736 to 0.743 | Improved |
| add background puzzle loss to reconstruction loss (ablation) | EraseLoRA · OpenImages V7 (200 samples) | FG Sim.: 0.158 to 0.151 | Improved |
| add reconstruction loss to puzzle loss (ablation) | EraseLoRA · OpenImages V7 (200 samples) | BG Sim.: 0.561 to 0.743 | Improved |
| use InternVL3-78B instead of LLaVA-7B as MLLM (ablation) | EraseLoRA · OpenImages V7 (200 samples) | BG Sim.: 0.728 to 0.743 | Mixed |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- + DesignEdit — DesignEdit: Unify Spatial-Aware Image Editing via Training-free Inpainting with a Multi-Layered Latent Diffusion Framework (AAAI 2025) · better on 5 of 5 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Precise Object and Effect Removal with Adaptive Target-Aware Attention (CVPR 2026 · similarity 0.88)
- Attentive Eraser: Unleashing Diffusion Model’s Object Removal Potential via Self-Attention Redirection Guidance (AAAI 2025 · similarity 0.86)
- SmartEraser: Remove Anything from Images using Masked-Region Guidance (CVPR 2025 · similarity 0.86)
- Detail-Preserving Latent Diffusion for Stable Shadow Removal (CVPR 2025 · similarity 0.85)
- In-N-Out: Lifting 2D Diffusion Prior for 3D Object Removal via Tuning-Free Latents Alignment (NeurIPS 2024 · similarity 0.85)
- CLIPAway: Harmonizing focused embeddings for removing objects via diffusion models (NeurIPS 2024 · 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.
