Research map / 3D vision / Depth and stereo
Defocus Blur Detection via Depth Distillation
Xiaodong Cun, Chi-Man Pun
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 depth distillation (ablation) | FCN with ResNeXt101 · DUT500 | F-beta: 0.879 -> 0.883 | Improved |
| add depth distillation (ablation) | FCN with ResNeXt101 · CUHK100 | F-beta: 0.917 -> 0.922 | Improved |
| add depth distillation (ablation) | FCN with ResNeXt101 · DUT500 | MAE: 0.080 -> 0.077 | Improved |
| add Selective Reception Field Block (ablation) | FCN with ResNeXt101 + depth · DUT500 | F-beta: 0.883 -> 0.888 | Improved |
| add Selective Reception Field Block (ablation) | FCN with ResNeXt101 + depth · CUHK100 | F-beta: 0.922 -> 0.926 | Improved |
| add selective attention to RFB (SRFB instead of FRB) (ablation) | FCN with ResNeXt101 + depth · CUHK100 | F-beta: 0.931 -> 0.927 | Hurt |
| add selective attention to RFB (SRFB instead of FRB) (ablation) | FCN with ResNeXt101 + depth · DUT500 | F-beta: 0.887 -> 0.891 | Improved |
| add Supervision-guided Attention Block (ablation) | FCN with ResNeXt101 · DUT500 | MAE: 0.076 -> 0.073 | Improved |
| switch backbone from VGG19 to ResNeXt101 | FCN · DUT500 | F-beta: 0.800 -> 0.879 | Improved |
| switch backbone from VGG19 to ResNeXt101 | FCN · CUHK100 | F-beta: 0.911 -> 0.917 | Improved |
| add depth distillation, SRFB and SAB on VGG19 (ablation) | FCN VGG · DUT500 | F-beta: 0.800 -> 0.844 | Improved |
| set depth distillation weight to 0.1 rather than 0.01 or 5 (ablation) | Ours · CUHK100 | F-beta: 0.9253 -> 0.9267 | Improved |
Closest papers by meaning
- Defocus Blur Detection via Multi-Stream Bottom-Top-Bottom Fully Convolutional Network (CVPR 2018 · similarity 0.94)
- DeFusionNET: Defocus Blur Detection via Recurrently Fusing and Refining Multi-Scale Deep Features (CVPR 2019 · similarity 0.94)
- DeFusionNET: Defocus Blur Detection via Recurrently Fusing and Refining Discriminative Multi-Scale Deep Features (TPAMI 2020 · similarity 0.93)
- Blurry-Edges: Photon-Limited Depth Estimation from Defocused Boundaries (CVPR 2025 · similarity 0.89)
- Multi-Focus Image Fusion via Explicit Defocus Blur Modelling (AAAI 2025 · similarity 0.89)
- Quad-Pixel Image Defocus Deblurring: A New Benchmark and Model (CVPR 2025 · similarity 0.89)
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.
