Research map / Learning theory and algorithms / Optimal transport
Lightweight Optimal-Transport Harmonization on Edge Devices
Maria Larchenko, Dmitry Guskov, Alexander Lobashev, Georgy Derevyanko
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 L1 instead of L2 loss on MKL filter labels (ablation) | MKL-Harmonizer · iHarmony4 test (256x256) | MSE: 66.3 -> 65.0 | No clear effect |
| predict [A, S] instead of [mu1, Sigma1] (ablation) | MKL-Harmonizer · iHarmony4 | training loss | Improved |
| add content L1 loss to label loss (ablation) | MKL-Harmonizer · iHarmony4 | filter collapse to identity | Improved |
| MKL-Harmonizer vs PCT-Net on in-distribution benchmark | MKL-Harmonizer · iHarmony4 test (256x256) | MSE: 29.1 -> 65.0 | Hurt |
| MKL-Harmonizer vs best baselines on real AR composites (human preference) | MKL-Harmonizer · ARCore user study | votes: 172 vs INR 200, PCT-Net 166 | No clear effect |
| switch from dense predictors to filter-based MKL at high resolution | harmonization · 4096x4096 inference speed | iterations per second: 11.84 (PCT-Net) -> 40.85 | Improved |
Closest papers by meaning
- Learning Image Harmonization in the Linear Color Space (ICCV 2023 · similarity 0.84)
- Progressive Painterly Image Harmonization from Low-Level Styles to High-Level Styles (AAAI 2024 · similarity 0.83)
- High-Resolution Image Harmonization with Adaptive-Interval Color Transformation (NeurIPS 2024 · similarity 0.82)
- Painterly Image Harmonization by Learning from Painterly Objects (AAAI 2024 · similarity 0.82)
- FRIH: Fine-Grained Region-Aware Image Harmonization (AAAI 2024 · similarity 0.82)
- Efficient Depth Estimation for Unstable Stereo Camera Systems on AR Glasses (CVPR 2025 · similarity 0.82)
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.
