Research map / 3D vision / 3D generation
Leveraging Multi-View Image Sets for Unsupervised Intrinsic Image Decomposition and Highlight Separation
Renjiao Yi, Ping Tan, Stephen Lin
This paper tried a common change: Add a contrastive loss
add contrastive loss for joint finetuning · ShapeNet and real · qualitative Improved
What 1,317 accepted papers report for this change
- 1,253 improved
- 22 no clear effect
- 16 mixed
- 26 hurt
- Beyond Looking Up, Try Looking Around: Harmonizing Global Structure and Local Consistency in Optimal Transport for Short Text Clustering ICML 2026 · add contrastive learning · SearchSnippets · ACC: 56.65 -> 80.12Improved
- SleepLM: Natural-Language Intelligence for Human Sleep ICML 2026 · add contrastive objective to captioning (CoCa vs Cap) · sleep event IoU · IoU: 29.2 -> 27.4Hurt
- LIMSSR: LLM-Driven Sequence-to-Score Reasoning under Training-Time Incomplete Multimodal Observations ICML 2026 · add contrastive loss L_con · FS1000 average · Spearman correlation: 0.759 -> 0.789Improved
See this change in detailMatched by the wording of each paper's own result, read by AI from the full text. Papers report gains more often than failures, so this shows what gets published, not the chance a change works.
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 unsupervised finetuning on real multi-view photos (ablation) | Highlight separation network · Real lab dataset | MSE (diffuse): 0.0045 -> 0.0036 | Improved |
| add unsupervised finetuning on real photos (ablation) | Highlight separation network · ShapeNet Intrinsic synthetic | MSE (diffuse): 0.0015 -> 0.0016 | Hurt |
| use color distribution loss instead of pixel-to-pixel low-rank loss (ablation) | Highlight separation network · Real lab dataset | MSE (diffuse): 0.0041 -> 0.0036 | Improved |
| use color distribution loss instead of pixel-to-pixel low-rank loss (ablation) | Highlight separation network · ShapeNet Intrinsic synthetic | MSE (diffuse): 0.0020 -> 0.0016 | Improved |
| add unsupervised finetuning (ablation) | Intrinsic decomposition network · ShapeNet Intrinsic | MSE (albedo): 0.0108 -> 0.0054 | Improved |
| use color distribution loss instead of pixel-to-pixel low-rank loss (ablation) | Intrinsic decomposition network · ShapeNet Intrinsic | MSE (shading): 0.0087 -> 0.0045 | Improved |
| use color distribution loss instead of pixel-to-pixel low-rank loss (ablation) | Intrinsic decomposition network · ShapeNet Intrinsic | MSE (albedo): 0.0067 -> 0.0054 | Improved |
| add contrastive loss for joint finetuning (ablation) | Full H-Net + S-Net · ShapeNet and real | qualitative | Improved |
| add supervised pretraining on 1.1% ShapeNet synthetic data (ablation) | Fully unsupervised network · Real lab dataset | MSE (diffuse): 0.0057 -> 0.0036 | Improved |
| add pretraining on synthetic data (ablation) | Fully unsupervised network · ShapeNet Intrinsic | MSE (albedo): 0.0067 -> 0.0054 | Improved |
| skip connections between encoder and decoder in H-Net (ablation) | H-Net · ShapeNet | qualitative | No clear effect |
Closest papers by meaning
- Unsupervised Learning for Intrinsic Image Decomposition From a Single Image (CVPR 2020 · similarity 0.88)
- Learning to Separate Multiple Illuminants in a Single Image (CVPR 2019 · similarity 0.86)
- Self-Supervised Intrinsic Image Decomposition (NeurIPS 2017 · similarity 0.86)
- Unsupervised Intrinsic Image Decomposition Using Internal Self-Similarity Cues (TPAMI 2021 · similarity 0.86)
- Generative Multiview Relighting for 3D Reconstruction under Extreme Illumination Variation (CVPR 2025 · similarity 0.85)
- Few-Shot Neural Radiance Fields under Unconstrained Illumination (AAAI 2024 · similarity 0.84)
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.
