Research map / Vision-language and multimodal / Multimodal LLMs
Lumina-OmniLV: A Unified Multimodal Framework for General Low-Level Vision
Yuandong Pu, Le Zhuo, Kaiwen Zhu, Liangbin Xie, Wenlong Zhang, Xiangyu Chen, Peng Gao, Yu Qiao, Chao Dong, Yihao 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 |
|---|---|---|---|
| use OmniLV instead of GenLV on deblurring | GenLV · OLV-T Deblur (6 types) | PSNR: 22.15 -> 22.57 | Improved |
| use OmniLV instead of PixWizard on depth estimation | PixWizard · Depth estimation | RMSE: 0.941 -> 0.525 | Improved |
| use OmniLV instead of PixWizard on HED edge detection | PixWizard · HED edge detection | MAE: 25.95 -> 20.75 | Improved |
| use OmniLV instead of specialized Depth Anything on depth | Depth Anything · Depth estimation | RMSE: 0.291 -> 0.525 | Hurt |
| use OmniLV instead of MAXIM on low-light enhancement | MAXIM · LOLv2-Real Low light | PSNR: 34.04 -> 18.60 | Hurt |
| use OmniLV instead of GenLV on stylization (LLF) | GenLV · LLF stylization | FID: 29.53 -> 24.16 | Improved |
| use OmniLV instead of GenLV on stylization (LLF) | GenLV · LLF stylization | PSNR: 25.66 -> 21.72 | Hurt |
| use OmniLV instead of PixWizard in human preference | PixWizard · Human preference benchmarks | Elo rating: 1,310 -> 1,475 | Improved |
| use OmniLV instead of PixWizard on dense task success rate (Image2Depth) | PixWizard · Dense prediction success rate | success rate: 93% -> 100% | Improved |
| separate text and visual encoding instead of unified MLLM encoding (ablation) | OmniLV · Dense prediction tasks | task alignment | Improved |
| fine-tune base model with adapter (co-training) instead of adapter only (ablation) | OmniLV · Condition integration settings | performance | Improved |
| use projection-addition instead of input concatenation for visual prompts (ablation) | OmniLV · In-context visual prompts | performance | Improved |
Closest papers by meaning
- Divide, Conquer and Combine: A Training-Free Framework for High-Resolution Image Perception in Multimodal Large Language Models (AAAI 2025 · similarity 0.88)
- Lumen: Unleashing Versatile Vision-Centric Capabilities of Large Multimodal Models (NeurIPS 2024 · similarity 0.88)
- LL-Bench: Rethinking Low-Level Vision Evaluation in the Era of Large-Scale Generative Models (NeurIPS 2026 · similarity 0.88)
- DeepStack: Deeply Stacking Visual Tokens is Surprisingly Simple and Effective for LMMs (NeurIPS 2024 · similarity 0.88)
- UVU: Improving Multimodal Understanding via Vision-Language Unified Autoregressive Paradigm (CVPR 2026 · similarity 0.88)
- iLLaVA: An Image is Worth Fewer Than 1/3 Input Tokens in Large Multimodal Models (ICLR 2026 · similarity 0.88)
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.
