Research map / 3D vision / 3D generation
IRISformer: Dense Vision Transformers for Single-Image Inverse Rendering in Indoor Scenes
Rui Zhu, Zhengqin Li, Janarbek Matai, Fatih Porikli, Manmohan Chandraker
Check 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 |
|---|---|---|---|
| Use single-task rather than multi-task models | IRISformer · IIW | WHDR: 13.1 -> 12.0 | Improved |
| Use single-task rather than multi-task models | IRISformer · NYUv2 | Mean normal error (degrees): 23.5 -> 20.2 | Improved |
| Apply the bilateral solver (ablation) | IRISformer single-task · OpenRooms | A error: 0.43 -> 0.43 | No clear effect |
| Apply the bilateral solver (ablation) | IRISformer single-task · OpenRooms | R error: 5.50 -> 5.48 | Improved |
| Apply the bilateral solver (ablation) | IRISformer single-task · OpenRooms | L+I error: 12.14 -> 12.17 | Hurt |
| Use six rather than four transformer layers (ablation) | IRISformer single-task · OpenRooms | Albedo loss: 0.48 -> 0.43 | Improved |
| Use six rather than four transformer layers (ablation) | IRISformer single-task · OpenRooms | Normal loss: 1.89 -> 1.89 | No clear effect |
| Use batch normalization in BRDFGeoNet output heads (ablation) | IRISformer · OpenRooms single-task albedo | Albedo loss: 0.51 -> 0.43 | Improved |
| Use batch normalization in BRDFGeoNet output heads (ablation) | IRISformer · OpenRooms single-task normals | Normal loss: 1.92 -> 1.89 | Improved |
| Omit batch normalization from LightNet heads (ablation) | IRISformer · OpenRooms lighting | Lighting loss: 13.23 -> 12.54 | Improved |
| Share rather than separate decoders for albedo and roughness (ablation) | IRISformer multi-task · OpenRooms albedo+roughness | Albedo loss: 0.51 -> 0.50 | Improved |
| Share rather than separate decoders for albedo and normals (ablation) | IRISformer multi-task · OpenRooms albedo+normals | Normal loss: 1.85 -> 1.88 | Hurt |
Closest papers by meaning
- IRIS: Inverse Rendering of Indoor Scenes from Low Dynamic Range Images (CVPR 2025 · similarity 0.91)
- Latent Intrinsics Emerge from Training to Relight (NeurIPS 2024 · similarity 0.87)
- Factored-NeuS: Reconstructing Surfaces, Illumination, and Materials of Possibly Glossy Objects (CVPR 2025 · similarity 0.87)
- Eclipse: Disambiguating Illumination and Materials using Unintended Shadows (CVPR 2024 · similarity 0.87)
- Generative Multiview Relighting for 3D Reconstruction under Extreme Illumination Variation (CVPR 2025 · similarity 0.87)
- Twinner: Shining Light on Digital Twins in a Few Snaps (CVPR 2025 · similarity 0.87)
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.
