Research map / Image generation / Editing and style transfer
Deep Image Harmonization with Globally Guided Feature Transformation and Relation Distillation
Li Niu, Linfeng Tan, Xinhao Tao, Junyan Cao, Fengjun Guo, Teng Long, Liqing Zhang
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 globally guided feature transformation (GIFT) to last decoder block (ablation) | iSSAM-style UNet · iHarmony4 (All) | MSE: 24.64 -> 21.39 | Improved |
| add GIFT to additional decoder or encoder blocks beyond D3 (ablation) | GIFT · iHarmony4 (All) | MSE: 21.39 -> 20.50 | No clear effect |
| modulate conv weights with global bottleneck feature instead of fixed learnable scales (ablation) | GIFT on D3 · iHarmony4 (All) | MSE: 24.33 -> 21.39 | Improved |
| use global feature rather than the local feature map as guidance (ablation) | GIFT on D3 · iHarmony4 (All) | MSE: 24.21 -> 21.39 | Improved |
| add relation distillation on encoder feature maps (ablation) | GIFT on D2,D3 and encoder · iHarmony4 (All) | MSE: 20.50 -> 19.46 | Improved |
| also apply relation distillation to decoder feature maps (ablation) | GiftNet · iHarmony4 (All) | MSE: 19.46 -> 21.31 | Hurt |
| use relation distillation instead of feature distillation on encoder (ablation) | GiftNet · iHarmony4 (All) | MSE: 21.58 -> 19.46 | Improved |
| GiftNet vs prior best on iHarmony4 Hday2night | GiftNet · iHarmony4 (Hday2night) | MSE: 33.14 -> 38.28 | Hurt |
| GiftNet vs strongest baselines on ccHarmony | GiftNet · ccHarmony | MSE: 27.87 -> 24.55 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 2 on most shared results.
- DCCF — DCCF: Deep Comprehensible Color Filter Learning Framework for High-Resolution Image Harmonization (ECCV 2022) · better on 3 of 3 shared results
- Harmonizer — Harmonizer: Learning to Perform White-Box Image and Video Harmonization (ECCV 2022) · better on 0 of 1 shared result
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
- Transformer for Image Harmonization and Beyond (TPAMI 2022 · similarity 0.90)
- FRIH: Fine-Grained Region-Aware Image Harmonization (AAAI 2024 · similarity 0.88)
- Progressive Painterly Image Harmonization from Low-Level Styles to High-Level Styles (AAAI 2024 · similarity 0.87)
- Relightful Harmonization: Lighting-aware Portrait Background Replacement (CVPR 2024 · similarity 0.87)
- Painterly Image Harmonization by Learning from Painterly Objects (AAAI 2024 · similarity 0.86)
- Training-and-Prompt-Free General Painterly Harmonization via Zero-Shot Disentenglement on Style and Content References (AAAI 2025 · similarity 0.86)
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.
