Research map / Label-efficient and robust learning / Domain adaptation and generalization
A Simple yet Mighty Hartley Diffusion Versatilist for Generalizable Dense Vision Tasks
Qi Bi, Jingjun Yi, Huimin Huang, Hao Zheng, Haolan Zhan, Wei Ji, Yawen Huang, Yuexiang Li, Yefeng Zheng
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 low-frequency training process (LTP) to DDP baseline (ablation) | DDP baseline · DGSS trained on SYNTHIA (C, B, M) | mIoU: 1.5%, 1.7% and 1.3% | Improved |
| add high-frequency sampling process (HSP) to DDP baseline (ablation) | DDP baseline · DGSS trained on SYNTHIA (C, B, M) | mIoU: 0.9%, 1.2% and 0.5% | Improved |
| add both LTP and HSP to DDP baseline (ablation) | DDP baseline · DGSS trained on SYNTHIA (avg) | mIoU: 54.7 -> 57.8 | Improved |
| inject V[0,10%) in training only (ablation) | HarDiff · DGSS trained on SYNTHIA (avg) | mIoU: 54.7 -> 56.2 | Improved |
| inject V[0,10%) in both training and sampling (ablation) | HarDiff · DGSS trained on SYNTHIA (avg) | mIoU: 54.7 -> 56.8 | Improved |
| inject V[90%,100%) in sampling only (ablation) | HarDiff · DGSS trained on SYNTHIA (avg) | mIoU: 54.7 -> 55.6 | Improved |
| inject V[90%,100%) in both training and sampling (ablation) | HarDiff · DGSS trained on SYNTHIA (avg) | mIoU: 54.7 -> 56.1 | Improved |
| inject both bands in both training and sampling instead of low in training and high in sampling (ablation) | HarDiff · DGSS trained on SYNTHIA (avg) | mIoU: 57.8 -> 57.6 | Hurt |
| use DHT instead of FFT, DCT or Haar wavelet (ablation) | HarDiff · DGSS S to C, B, M | mIoU | Improved |
| remove highest-frequency band V[0,10%) in band-rejection analysis (ablation) | ConvNeXt binary classifier · Trained on RESIDE, source/target accuracy | accuracy | Hurt |
Closest papers by meaning
- Exploiting Diffusion Prior for Generalizable Dense Prediction (CVPR 2024 · similarity 0.89)
- Lotus: Diffusion-based Visual Foundation Model for High-quality Dense Prediction (ICLR 2025 · similarity 0.88)
- What Matters When Repurposing Diffusion Models for General Dense Perception Tasks? (ICLR 2025 · similarity 0.88)
- Scaling Properties of Diffusion Models For Perceptual Tasks (CVPR 2025 · similarity 0.88)
- Diff-2-in-1: Bridging Generation and Dense Perception with Diffusion Models (ICLR 2025 · similarity 0.87)
- Unleashing Diffusion Transformers for Visual Correspondence by Modulating Massive Activations (NeurIPS 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.
