AI Paper Atlas

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

CVPR 2022 · 3D vision › 3D generation · approach: Generative indoor scene synthesis · 48 citations (OpenAlex)

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 changedWhereMeasuredEffect
Use single-task rather than multi-task modelsIRISformer · IIWWHDR: 13.1 -> 12.0Improved
Use single-task rather than multi-task modelsIRISformer · NYUv2Mean normal error (degrees): 23.5 -> 20.2Improved
Apply the bilateral solver (ablation)IRISformer single-task · OpenRoomsA error: 0.43 -> 0.43No clear effect
Apply the bilateral solver (ablation)IRISformer single-task · OpenRoomsR error: 5.50 -> 5.48Improved
Apply the bilateral solver (ablation)IRISformer single-task · OpenRoomsL+I error: 12.14 -> 12.17Hurt
Use six rather than four transformer layers (ablation)IRISformer single-task · OpenRoomsAlbedo loss: 0.48 -> 0.43Improved
Use six rather than four transformer layers (ablation)IRISformer single-task · OpenRoomsNormal loss: 1.89 -> 1.89No clear effect
Use batch normalization in BRDFGeoNet output heads (ablation)IRISformer · OpenRooms single-task albedoAlbedo loss: 0.51 -> 0.43Improved
Use batch normalization in BRDFGeoNet output heads (ablation)IRISformer · OpenRooms single-task normalsNormal loss: 1.92 -> 1.89Improved
Omit batch normalization from LightNet heads (ablation)IRISformer · OpenRooms lightingLighting loss: 13.23 -> 12.54Improved
Share rather than separate decoders for albedo and roughness (ablation)IRISformer multi-task · OpenRooms albedo+roughnessAlbedo loss: 0.51 -> 0.50Improved
Share rather than separate decoders for albedo and normals (ablation)IRISformer multi-task · OpenRooms albedo+normalsNormal loss: 1.85 -> 1.88Hurt

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy