Research map / Image generation / Text-to-image diffusion
UniPercept: A Unified Diffusion Model for Generalizable Visual Perception
Zuyan Zhao, Zhenliang He, Meina Kan, Shiguang Shan, Xilin Chen
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 |
|---|---|---|---|
| train foundation on all 7 base tasks instead of depth only (ablation) | UniPercept · Depth | AbsRel: 8.4 -> 7.3 | Improved |
| train foundation on all 7 base tasks instead of normal only (ablation) | UniPercept · Normal | mean angular error: 18.5 -> 17.4 | Improved |
| add more base tasks (depth only to all 7) (ablation) | foundation model · DIS | max F-measure: 0.746 -> 0.752 | Improved |
| add more base tasks (depth only to all 7) (ablation) | foundation model · BSDS500 edge | ODS: 0.766 -> 0.773 | Improved |
| use foundation-adapter model instead of Sana baseline for DIS adaptation (ablation) | UniPercept · DIS | training steps to converge: over 30,000 -> 1000 steps | Improved |
| widen adapter bottleneck (ratio 128 to 16) (ablation) | adapter · metallicity | PSNR and SSIM: not given numerically | Improved |
| fine-tune with 1K samples instead of full data (ablation) | UniPercept adapter · BSDS500 edge | ODS: 0.806 -> 0.810 | No clear effect |
| fine-tune with 1K samples instead of full data (ablation) | UniPercept adapter · DIS5K | max F-measure: .764 -> .758 | No clear effect |
| use UniPercept instead of specialist BiRefNet | DIS adapter · DIS5K | max F-measure: .896 -> .764 | Hurt |
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.
- PixWizard — PixWizard: Versatile Image-to-Image Visual Assistant with Open-Language Instructions (ICLR 2025) · better on 2 of 2 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
- DICEPTION: A Generalist Diffusion Model for Visual Perceptual Tasks (NeurIPS 2025 · similarity 0.91)
- Scaling Properties of Diffusion Models For Perceptual Tasks (CVPR 2025 · similarity 0.89)
- DEEM: Diffusion models serve as the eyes of large language models for image perception (ICLR 2025 · similarity 0.89)
- UniPercept: Towards Unified Perceptual-Level Image Understanding across Aesthetics, Quality, Structure, and Texture (ICML 2026 · similarity 0.88)
- Diff-2-in-1: Bridging Generation and Dense Perception with Diffusion Models (ICLR 2025 · similarity 0.88)
- What Matters When Repurposing Diffusion Models for General Dense Perception Tasks? (ICLR 2025 · 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.
