Research map / Image generation / Autoregressive and latent generation
ShadingNet: Image Intrinsics by Fine-Grained Shading Decomposition
Anil S. Baslamisli, Partha Das, Hoàng-Ân Lê, Sezer Karaoğlu, Theo Gevers
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 refinement module fusing branch reflectances (ablation) | ShadingNet · NED test | SMSE: 0.0030 -> 0.0027 | Improved |
| add refinement module fusing branch reflectances (ablation) | ShadingNet · GTA V | SMSE: 0.0125 -> 0.0124 | Improved |
| add refinement module (ablation) | ShadingNet · NED test | DSSIM: 0.0910 -> 0.0798 | Improved |
| explicitly classify shadow casts in a dedicated branch instead of unified shading (ablation) | ShadingNet · NED test | SMSE: 0.0030 -> 0.0030 (LMSE 0.0157 -> 0.0144) | Improved |
| add fine-grained shading decomposition to ShapeNet backbone (Baseline-a) | ShapeNet · NED test | SMSE albedo: 0.0075 -> 0.0072 | Improved |
| add fine-grained shading decomposition to ShapeNet backbone (Baseline-b) | ShapeNet · GTA V | SMSE: 0.0138 -> 0.0134 | No clear effect |
| use ShadingNet design instead of naive fine-grained baselines | Baseline-a · NED test | SMSE albedo: 0.0072 -> 0.0027 | Improved |
| use individual decoders (Baseline-b) rather than single unified shading decoder (Baseline-a) | fine-grained baselines · NED direct shading | SMSE: 0.0545 -> 0.0579 | Hurt |
| use individual decoders (Baseline-b) rather than single unified decoder (Baseline-a) | fine-grained baselines · IIW outdoor | WHDR: 46.22 -> 39.11 | Improved |
| apply guided filter post-processing | ShadingNet · IIW outdoor | WHDR: 35.73 -> 29.98 | Improved |
| compare with image-sequence-trained IIDWW using smoothness priors | ShadingNet · IIW outdoor | WHDR: 21.60 vs 35.73 | Hurt |
Closest papers by meaning
- Intrinsic Image Transformation via Scale Space Decomposition (CVPR 2018 · similarity 0.90)
- Unsupervised Learning for Intrinsic Image Decomposition From a Single Image (CVPR 2020 · similarity 0.90)
- Unsupervised Intrinsic Image Decomposition Using Internal Self-Similarity Cues (TPAMI 2021 · similarity 0.90)
- Revisiting Deep Intrinsic Image Decompositions (CVPR 2018 · similarity 0.90)
- Self-Supervised Intrinsic Image Decomposition (NeurIPS 2017 · similarity 0.90)
- When Shadow Removal Meets Intrinsic Image Decomposition: A Joint Learning Framework Using Unpaired Data (AAAI 2025 · similarity 0.85)
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.
