Research map / Image generation / Editing and style transfer
EditCLIP: Representation Learning for Image Editing
Qian Wang, Aleksandar Cvejić, Abdelrahman Eldesokey, Peter Wonka
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 |
|---|---|---|---|
| use EditCLIP edit embedding instead of InstaManip conditioning | exemplar-based editing · TOP-Bench-X | EditCLIP-to-EditCLIP: 0.383 -> 0.477 | Improved |
| use EditCLIP instead of VISII | exemplar-based editing · TOP-Bench-X | runtime (s): 370 -> 1.8 | Improved |
| use EditCLIP instead of InstaManip | exemplar-based editing · TOP-Bench-X | runtime (s): 14.9 -> 1.8 | Improved |
| use EditCLIP instead of InstaManip | exemplar-based editing · TOP-Bench-X | S_visual: 0.735 -> 0.761 | Hurt |
| use EditCLIP instead of PromptDiffusion | exemplar-based editing · TOP-Bench-X | LPIPS: 0.620 -> 0.233 | Improved |
| use joint input/edited-image encoding instead of embedding only the edited image with CLIP (ablation) | EditCLIP exemplar editing · TOP-Bench-X | qualitative edit transfer | Improved |
| use joint EditCLIP encoding instead of summing, channel-concatenating or sequence-appending CLIP embeddings (ablation) | EditCLIP exemplar editing · TOP-Bench-X | qualitative edit transfer | Improved |
| use last-layer hidden states instead of projected embedding (ablation) | EditCLIP exemplar editing · TOP-Bench-X | edit transfer and layout preservation | Improved |
| use ViT-L/14 backbone instead of ViT-B/32 (ablation) | EditCLIP · exemplar editing qualitative | detail preservation | Mixed |
| add input preservation LPIPS loss (lambda2 0.05) (ablation) | EditCLIP exemplar editing · TOP-Bench-X | layout preservation vs edit strength: 0 -> 0.05 | Mixed |
| use EditCLIP-to-Text instead of CLIP directional similarity as human-aligned metric | edit evaluation · TOP-Bench-X | Pearson correlation edit quality: 0.186 -> 0.256 | Improved |
| use EditCLIP-to-EditCLIP instead of S_visual | edit evaluation · TOP-Bench-X | Pearson correlation edit quality: 0.240 -> 0.372 | Improved |
Closest papers by meaning
- Preserve or Modify? Context-Aware Evaluation for Balancing Preservation and Modification in Text-Guided Image Editing (CVPR 2025 · similarity 0.89)
- Instruct-CLIP: Improving Instruction-Guided Image Editing with Automated Data Refinement Using Contrastive Learning (CVPR 2025 · similarity 0.89)
- PairEdit: Learning Semantic Variations for Exemplar-based Image Editing (NeurIPS 2025 · similarity 0.89)
- ImgEdit: A Unified Image Editing Dataset and Benchmark (NeurIPS 2025 · similarity 0.88)
- Emu Edit: Precise Image Editing via Recognition and Generation Tasks (CVPR 2024 · similarity 0.87)
- Concept Lancet: Image Editing with Compositional Representation Transplant (CVPR 2025 · similarity 0.87)
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.
